Information doi: 10.3390/info17080770

Authors: Andreas Triantafyllidis Sofia Segkouli Evdoxia Eirini Lithoxoidou Anastasios Alexiadis Konstantinos Votis Kleio Koutra Vassilis Kilintzis Haridimos Kondylakis Eunate Arana-Arri Severin Haug Nikolaos Boumparis Maria Krini Liselot Hudders Dimitrios Tzovaras

Objective: Large Language Models (LLMs) are emerging as a key component for digital health systems based on Artificial Intelligence (AI). The objective of this paper is to systematically review the characteristics, effectiveness, and implementation challenges of LLM-based assistants targeting adolescent health and well-being. Methods: A systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. PubMed, Scopus, and Web of Science were searched in November 2025 for studies published from 2022 onwards. Eligible studies examined LLM-based assistants used directly by adolescents in health and well-being contexts and reported quantitative outcomes. Data was extracted independently by multiple reviewers and synthesised narratively. Risk of bias was assessed using the Mixed Methods Appraisal Tool (MMAT). Results: Nine studies met the inclusion criteria, involving between 3 and 40 participants and covering mental health support, physical activity promotion, treatment engagement, vaccination awareness, and academic self-efficacy. Most studies were proof-of-concept investigations or pilot studies. LLM-based assistants were associated with reductions in depression, anxiety, stress, and negative affect, improvements in emotional regulation, physical activity, treatment motivation, HPV knowledge, and academic self-efficacy. Risk of bias was generally moderate and the evidence base was limited by small samples, short follow-up periods, and heterogeneous methodologies. Conclusions: LLM-based assistants show promise as scalable and accessible tools for supporting adolescent health and well-being, particularly in mental health domains. However, the current evidence remains preliminary. Larger, long-term, and rigorously designed studies are required to establish effectiveness, safety, and implementation feasibility. Registration: This review was not pre-registered and no review protocol has been published prior to conducting the review.

Information doi: 10.3390/info17080769

Authors: Francesco Benedetto Federica Massimi

Profiling side-channel attacks based on deep learning can recover cryptographic information from power-consumption traces, but their effectiveness may be reduced when desynchronisation displaces informative leakage across temporal positions. This study investigates whether Temporal Convolutional Networks (TCNs) can model the local and long-range dependencies present in desynchronised traces. Three complementary TCN variants are designed to isolate different temporal modelling strategies: a single-kernel dilated architecture (TCN1), a multi-scale architecture using parallel kernel sizes (TCN2), and a residual architecture intended to support stable hierarchical feature learning (TCN3). The models are evaluated on the ASCAD v1 fixed-key benchmark under synchronised conditions and maximum temporal shifts of 25, 50, and 75 samples, using validation loss and complementary key-ranking metrics. Under the most challenging setting, TCN1 achieves the highest Rank Success Rate and reaches its best rank substantially earlier than the conventional CNN baselines, although the lowest Final Rank is obtained by a CNN. These results indicate that TCN1 provides the most favourable trade-off among convergence speed, ranking consistency, and architectural complexity, without establishing uniform TCN superiority across all metrics. Gradient saliency, LIME, and occlusion analyses further identify temporal regions that influence the model predictions and are consistent with expected leakage patterns. The main contribution is a controlled comparison of complementary TCN design strategies, combined with an explainability analysis, for profiling side-channel attacks under trace desynchronisation.

Information doi: 10.3390/info17080768

Authors: Aizhan Kassymova Raissa Uskenbayeva Young Im Cho Venera Elle Aizhan Anartayeva Aizhan Smakhanova

Group-buying platforms require the joint treatment of individual user relevance and hard operational constraints, including minimum group size, lot capacity, spending limits, product availability, and simultaneous participation limits. This paper presents Preference-Aware Hybrid Combinatorial Group Optimization (PA-HCGO), a prototype-calibrated decision framework that integrates implicit feedback preference learning with constrained user–lot assignment. The contribution is not a new recommender architecture or a new integer programming solver, but a system-level integration of learned user–lot utility and hard group-buying feasibility constraints inside the SmartBuy Connect prototype. The empirical part uses an anonymized prototype dataset containing 200 products, 150 users, 150 lots, 4000 user events, and 500 orders. The evaluation is explicitly divided into three tracks: observed temporal recommendation testing, a calibrated counterfactual pre-activation scenario, and synthetic scalability tests generated from prototype-calibrated distributions. In the observed transactional test subset, which contains 40 held-out transactional items from 36 users, the PA-PREF preference layer achieved Recall@10 = 0.3611 and NDCG@10 = 0.1319 using user–category profiles available at the end of the training interval. Although PA-PREF obtained the highest point estimate, its advantage over the content-only baseline was not statistically separable at the 95% level on this small transactional subset. The result is therefore treated as preliminary. For general engagement, the popularity baseline remained stronger, indicating that the proposed preference model is more useful for transactional intent prediction than for all activity types. In the calibrated counterfactual pre-activation scenario, PA-HCGO activated 117 of 150 lots, compared with 85 lots under independent greedy assignment. This scenario is a constructed pre-activation setting rather than an observed historical platform state. In synthetic scalability tests, the method remained computationally feasible up to 5000 users and 1000 lots, with a mean solve time of 4.7174 s in the prototype implementation. The results suggest that learned user–lot utility can improve constrained group formation, but the evidence should be interpreted as prototype-calibrated rather than as proof of industrial-scale business effectiveness.

Information doi: 10.3390/info17080767

Authors: Victor J. Vera Ponce Fiorella E. Zuzunaga-Montoya Jhosmer Ballena-Caicedo Juan Carlos Bustamante-Rodriguez Holly Estrella Delgado-Toro Carmen Ines Gutierrez De Carrillo

National health surveys generate reusable population-based evidence, but repeated use may produce thematic concentration, heterogeneous methodological reporting, and analytical overlap. This revised bibliometric and meta-research study audited 328 Scopus records retrieved for Peru’s Demographic and Family Health Survey (ENDES), excluded 14 records without eligible analytical ENDES use, and analyzed 314 publications. Matched full texts were available for 284 articles. Publications spanned 2001–2026 and increased mainly from 2019 onward. The most frequent thematic domains were Cardiometabolic health/adiposity (n = 93), Maternal/reproductive health (n = 77), and Child nutrition and growth (n = 39). The most reused ENDES years after validation were 2019 (n = 110), 2018 (n = 81), 2017 (n = 76), 2021 (n = 68), and 2022 (n = 66). Initial ENDES-year extraction showed a 25.4% false-positive rate, a 45.8% false-negative rate, and 28.3% complete article-level agreement. At a pairwise similarity threshold of 0.80, 32 analytical-overlap clusters were identified. The findings characterize the Scopus-indexed and Scopus-retrieved ENDES corpus and emphasize the need for explicit corpus auditing, transparent coding rules, validated survey-year extraction, and reproducible scripts.

Information doi: 10.3390/info17080766

Authors: Zhanna Sadirmekova Madina Sambetbayeva Bayangali Abdygalym Roman Taberkhan Anar Sultangaziyeva

Integrating newly emerging terms into existing ontologies is a recurring maintenance problem in knowledge engineering, particularly in biomedical domains where terminology evolves faster than manual curation can accommodate. This paper addresses the placement of concepts that are absent from the target ontology—the out-of-knowledge-base setting—in which a textual mention must be assigned one or more insertion positions in the subsumption hierarchy rather than linked to an existing node. We propose a three-stage framework that extends the conventional retrieve-then-select paradigm with an intermediate stage of edge generation and enrichment, which expands the candidate set by traversing the local structure of the ontology. Stage 1 retrieves candidate edges using a fine-tuned bi-encoder trained with a max-margin objective; Stage 2 constructs and structurally enriches candidate edges; Stage 3 selects among them using either a fine-tuned cross-encoder or a large language model under explainable instruction tuning. We evaluate on two datasets derived from SNOMED CT, MM-S14-Disease and MM-S14-CPP, under a strict out-of-knowledge-base protocol. Fine-tuned pre-trained language models outperform zero-shot and instruction-tuned large language models on ranking accuracy, while the instruction-tuned configuration produces expert-auditable justifications at a modest cost in accuracy. On MM-S14-Disease, the strongest configuration places a correct insertion edge among the ten highest-ranked candidates for 38.7% of test mentions and recovers the complete gold edge set for 16.4%, against 26.1% and 9.2% for retrieval alone. The framework is positioned as decision support for ontology curators rather than as an autonomous ontology generator.

Information doi: 10.3390/info17080765

Authors: Assel Ospan Madina Mansurova Zhansaya Zhangabay Aktoty Zhappar

Large language model (LLM)-based assistants can make large sociological survey collections more accessible, but they may also increase the exposure of respondent-level information and socially sensitive content. This study presents SocioTable-KZ, a privacy-aware, risk-reducing pipeline for bilingual natural-language interpretation of relational sociological survey data from Kazakhstan. The source collection contains 2,385,890 records distributed across 28 linked tables. The pipeline combines identifier exclusion, deterministic JoinGraph serialisation, language-specific QLoRA adaptation of Qwen-family models, a three-class Safe–Sensitive–Unsafe safety module, constrained rewriting, and controlled release. Reference analytical texts were prepared through expert-curated seed examples followed by few-shot candidate generation and factual review. Generation experiments used independent Kazakh and Russian 80/10/10 splits with seed 42. The best reported Qwen3-4B configuration achieved BLEU/ROUGE-L/chrF scores of 29.07/49.88/58.31 for Kazakh and 46.67/65.00/68.40 for Russian. The safety-labelled corpus contained 1347 Kazakh and 3376 Russian instances. On the reported Kazakh evaluation corpus, the safety module achieved 0.925 accuracy and F1-scores of 0.87, 0.95, and 0.70 for Safe, Sensitive, and Unsafe, respectively; Unsafe recall was 0.59. The revised deployment protocol therefore prohibits unattended respondent-level release and requires deterministic identifier screening and human review for non-Safe or uncertain outputs. An official letter from the original data-collecting institution confirms active informed consent and authorised scientific use of the anonymised dataset. The evidence supports preliminary feasibility, not a formal privacy guarantee or production-readiness claim.

Information doi: 10.3390/info17080764

Authors: Ionut Dancau Catalin Dumitrescu Stefan Vasian Eduard Popovici Mara Chiosea

Electronic attack (EA) using jamming signals is an essential component of electronic warfare (EW), consisting of the use of electromagnetic energy to disrupt, block or reduce the effectiveness of enemy communications, radar and navigation systems. In current conflicts, jamming EA is also successfully used against drones (air/ground), which have become an important component of modern warfare. In this context, the article proposes a method of protection against jamming (anti-jamming) for drone communications, thus achieving their resilience to electronic attack. The proposed method combines the wavelet transform with threshold SURE, adaptive filtering (LMS) and Bidirectional Long Short-Term Memory for anti-jamming resilience of 64-QAM (quadrature amplitude modulation) communication used by drones, evaluates the performance of the results obtained against electronic warfare systems and presents the open challenges for future research.

Information doi: 10.3390/info17080763

Authors: Chan Liu Junhao Li Qin Liao

Continuous-variablequantum key distribution (CVQKD) enables information-theoretically secure communications between two legitimate users. However, practical CVQKD systems remain vulnerable to various attacks, while existing machine-learning-based detection schemes suffer from increasing computational complexity when processing large-scale monitoring data. In this paper, we propose a quantum k-nearest neighbor (QkNN)-based multiclass attack detection framework for CVQKD systems. Specifically, physical features extracted from Bob’s monitoring data are encoded into quantum states and subsequently classified using the QkNN algorithm to identify different attack behaviors. Furthermore, by incorporating the attack ratio, retained-data ratio, and attack detection performance into secret key rate analysis, an attack-aware secret key rate model is established to characterize the influence of practical attacks on key generation. Simulation results demonstrate that the proposed scheme achieves high attack classification accuracy while significantly reducing computational complexity. Moreover, the proposed framework establishes a direct connection between attack detection and secret key generation, enabling a more realistic security evaluation for practical CVQKD systems.

Information doi: 10.3390/info17080762

Authors: Xudong Xie Zhimao Lu Nianmin Yao

Software vulnerability detection plays a critical role in improving software quality, system reliability, and security assurance. Prompt-based adaptation offers a lightweight alternative for vulnerability detection in low-resource Common Weakness Enumeration (CWE) settings, where labeled target-domain data are limited and model fine-tuning can be costly or unstable. This paper proposes an Integrated-Gradient-Guided Prompt Optimization (IGPO) method that uses attribution feedback to guide large language models in revising prompts for low-resource cross-CWE vulnerability detection. IGPO keeps the vulnerability detector fixed, evaluates the current prompt on target validation data, identifies false-positive and false-negative cases, computes Integrated Gradients (IG) for misclassified functions, aggregates token-level attributions into line-level feedback, and uses a large language model to diagnose error patterns and optimize the prompt. Experiments on C/C++ functions from PrimeVul, DiverseVul, and BigVul across 14 CWE categories show that, on positive-transfer source–target pairs, IGPO improves the average F1 score from 0.6780 to 0.7267 and outperforms the zero-shot baseline on 94.97% of them. These results position IGPO as an attribution-informed prompt adaptation framework that improves cross-CWE vulnerability detection without updating the detector, while highlighting the importance of backbone suitability under negative transfer. This method is particularly useful for security researchers and practitioners who need to adapt vulnerability detectors to new CWE categories with limited labeled data.

Information doi: 10.3390/info17080761

Authors: Lei Pan Tong Geng Yuheng Liu

Multimodal aspect-based sentiment analysis (MABSA) predicts the sentiment polarity associated with a specified aspect by jointly exploiting textual and visual information. Existing models may be sensitive to limited prompting examples, cross-modal noise, and unreliable generated context. This paper proposes QMPN, Quality-Aware Memory Prompting Network, that stores sample-specific prompts derived from a small support set, retrieves relevant prompting evidence for each query, and uses the retrieved prompts to guide aspect-aware context generation. A task-oriented quality gate, learned indirectly through the sentiment classification objective, controls the contribution of the generated context to the final prediction. Under the fixed protocol used in this study, QMPN employs 50 labeled support instances for prompt-memory construction and achieves 78.6% accuracy and 74.8% macro-F1 on Twitter-2015, and 72.0% accuracy and 70.5% macro-F1 on Twitter-2017. Relative to the variant without context generation, the complete model improves accuracy/macro-F1 by 3.31/2.83 percentage points on Twitter-2015 and 3.84/3.22 percentage points on Twitter-2017. Ablation and parameter analyses further show the contributions of memory retrieval, adaptive prompt selection, context generation, and quality-aware fusion. Because the evaluation is limited to two historical Twitter benchmarks and a single fixed seed, the reported results should be interpreted within this experimental scope rather than as evidence of universal cross-domain generalization.

Information doi: 10.3390/info17080760

Authors: Eleonora Koltsova Maksim Pysin Alexey Lobanov Anatoly Antipov Alexey Arkhipov Anton Perekatov Roman Krasheninnikov

Industrial digital twins integrate physical objects, dynamic models, control systems, data analysis tools, 2D and 3D visualization, and existing software systems. Much research has focused on the functional composition of the digital twin, modeling, and application scenarios, while the organization of the digital twin as an evolving software system composed of technologically heterogeneous and autonomous subsystems remains insufficiently formalized. The goal of this study is to develop a conceptual composite architecture for an industrial digital twin, in which complex subsystems are viewed as highly interconnected and loosely coupled service components of a higher-order system. The research method is based on analogy, transfer, and adaptation of proven principles of distributed and microservice systems to the constraints of industrial digital twins. An architectural model is proposed that includes a physical object, a process model, SCADA subsystems, a unified data exchange subsystem, 2D and 3D representations, VR/AR components, and automated model building modules. The practical feasibility of the approach is demonstrated using a proof-of-concept digital twin of a methanol–ammonia co-production plant, integrating Honeywell UniSim Design R460.1, web-based SCADA, Unity, and specialized 2D and 3D representation generation modules. This demonstration confirms the feasibility of integrating independently developed subsystems and the technological heterogeneity of the solution, but does not constitute a production test of performance, scalability, or cost effectiveness. Requirements for contract stability, a consistent interaction environment, assigned data responsibility, version compatibility, and complete documentation are defined. Research limitations and areas for subsequent quantitative architecture validation are identified.

Information doi: 10.3390/info17080759

Authors: Dilyara Nazyrova Zhangeldi Aitkozha Valery Starovoitov

Forests cover nearly one-third of the Earth’s land surface and are subject to increasing anthropogenic pressure, including unauthorised construction, infrastructure expansion, and habitat fragmentation. Detecting buildings concealed beneath forest canopy is essential for environmental monitoring and territorial surveillance, yet optical satellite imagery fails under persistent cloud cover and dense vegetation, and no existing approach provides reliable detection across contrasting forest ecosystems. We address this gap by proposing a hybrid SAR-based detection framework that combines twelve physically interpretable Scattering-Informed SAR Features (SISF)—derived from electromagnetic scattering theory across amplitude, polarimetric, temporal, and texture dimensions—with a universal Convolutional Neural Network, integrated through probability-level fusion with isotonic regional calibration. Rather than relying on individual feature thresholds, the framework identifies buildings through their characteristic multidimensional scattering signature—a combination that remains discriminative across biomes where any single SAR descriptor would fail. The framework was evaluated on a novel 7721-object multi-biome benchmark spanning forest-steppe (Kazakhstan), boreal forest (Komi Republic, Russia), and tropical rainforest (Brazil, Pará). The final Fusion + Regional Calibration model achieved an overall F1-score of 0.803 on the held-out test set (N = 1545), outperforming single-model baselines by up to 17 percentage points. Regional F1-scores ranged from 0.727 (Kazakhstan) to 0.944 (Komi Republic), with detection performance remaining robust under partial canopy occlusion (F1 = 0.911, versus 0.903 for unobscured buildings). An empirical inverse relationship between hard negative proportion and detection F1-score was identified within each biome, with the 28–35% proportions present in our sampled regions reported as a preliminary observation rather than a generally optimal range—a dataset design finding not previously reported in the literature. The proposed framework provides a physically interpretable solution for SAR-based building detection under forest canopy, demonstrating consistent performance across three contrasting forest biomes, with direct applications to environmental monitoring and territorial surveillance in forested regions.

Information doi: 10.3390/info17080758

Authors: Nursultan Kuldeyev Emil Andekin Vassiliy Serbin Darkhan Yerezhep Orazmukhamed Bekmurat Kanibek Sansyzbay Yelena Bakhtiyarova Laura Tasbolatova

This study presents a methodology for cross-ranking analysis to evaluate the consistency of universities’ positions across the QS World University Rankings, Times Higher Education (THE) World University Rankings, and Academic Ranking of World Universities (ARWU). Data for 2022–2025 were harmonized by university names, countries, years, and ranking positions, resulting in a unified dataset of 316 observations. Ranking agreement was assessed using the overlap of ranked lists, a Spearman distance–based similarity measure, and a normalized inverse-rank metric. QS rankings were treated as the response variable, while THE and ARWU positions served as predictors. Model performance was evaluated using leave-one-year-out cross-validation for a median baseline, the average of THE and ARWU ranks, linear regression, Random Forest, and XGBoost. Random Forest achieved the lowest mean absolute error (MAE = 24.80 ± 3.26 ranking positions), whereas linear regression best preserved relative ordering (Spearman’s ρ = 0.794 ± 0.062). Additional analyses of the TOP-50, TOP-100, and TOP-200 subsets showed that prediction accuracy decreases as ranking depth increases. The findings demonstrate that QS, THE, and ARWU provide complementary rather than interchangeable assessments of university performance. The proposed framework is intended for comparative cross-ranking analysis rather than forecasting future university rankings.

Information doi: 10.3390/info17080757

Authors: Cathrine Linnes Giulio Ronzoni Joseph Lema Babu George Jerome Agrusa

Generative artificial intelligence (GenAI) now produces synthetic text, images, audio, and video at a quality and cost that place convincing synthetic fabrication within reach of non-specialist users. Deepfakes, synthetic media that alter a person’s appearance, voice, or behavior through machine learning (ML), are one of the most contested applications of this capability, yet public willingness to accept them under regulation remains less understood. This study examines how perceived benefits, perceived risks, privacy concerns, and demographic characteristics relate to trust in deepfake technology under regulatory safeguards. Survey data from 924 respondents from several countries were analyzed using descriptive statistics, independent-samples t-tests, analysis of variance, multiple regression, and thematic analysis of open-ended responses. Respondents recognized the potential benefits of deepfake technology for digital content creation and education while expressing widespread concern about misinformation, privacy violations, and criminal misuse. When respondents evaluated deepfake technology under an assumed privacy-protecting regulatory scenario, perceived risks did not independently predict trust, while perceived benefits were the strongest predictors. The findings indicate that institutional confidence may contribute to public acceptance of beneficial applications of generative AI, although the cross-sectional design does not establish a causal effect of regulation.

Information doi: 10.3390/info17080756

Authors: Turdybek Kurmetkan Orken Mamyrbayev Adem Tekerek Ainur Toleu

This study presents an audio–visual speech recognition (AVSR) model for Kazakh that jointly exploits audio and visual channels. The study introduces QazAVSR, a 57 h dataset collected from 271 speakers, and extracts synchronized audio signals and lip-region video sequences using FFmpeg 7.0, Dlib 19.24, and OpenCV 4.9.0. The proposed architecture uses the self-supervised HuBERT_BASE model in the audio branch and an ImageNet-pretrained ViT-B/16 model in the visual branch. Audio and visual representations are fused by a three-layer BiModalHformer block, where intra- and cross-attention operations are performed at each level. Extensive experimental validation, supplemented by rigorous paired bootstrap resampling significance tests, demonstrates that the full multimodal BiModalHformer model achieves a highly robust average character error rate (CER) of 31.2% and a Word Error Rate (WER) of 43.1%. These results significantly outperform traditional audio-only, video-only, and standard representation-level fusion baselines. Furthermore, comparisons against powerful external baseline architectures—including Whisper-Small and AV-HuBERT configurations rigorously adapted for the Kazakh language—statistically validate the architectural efficacy of the BiModalHformer framework. Additional systematic evaluations utilizing extended metrics such as the Match Error Rate (MER), word information preserved (WIP), and the Multimodal Synergy Index (MSI) confirm that the full audio–visual configuration preserves lexical information significantly more effectively. Finally, extensive noise perturbation experiments confirm that the multimodal architecture exhibits superior structural robustness to complex acoustic distortions, including environmental noise, synthetic room reverberation, and overlapping speech topologies.

Information doi: 10.3390/info17080755

Authors: Nashat Ali Almasria Ayman Abu Haija Manaf Al-Okaily Ahmad Abu-Dawleh Riyad Neman Darwazeh Yasean A. Tahat

Information Systems (IS) play a critical role in managing and optimizing core business processes and financial operations within small and medium-sized enterprises (SMEs). Specifically, Cloud-based Accounting Information Systems (CAIS) represent a significant digital innovation that is reshaping accounting practices and financial management in SMEs. Despite growing adoption, limited empirical evidence exists on the organizational and governance factors influencing CAIS implementation in developing economies. This paper conducts empirical research primarily examining how organizational culture, organizational structure, and organizational support shape the effective adoption and usage of CAIS among SMEs in Jordan. In particular, the study explores the moderating role of Information Technology Governance (ITG) in the relationships between these organizational factors and CAIS usage. Drawing on the Information Systems and digital accounting literature, this study employs quantitative research design using PLS-SEM. An adapted model is derived from the Technology–Organization–Environment perspectives in adoption technological and contextual factors. Data was collected through a structured questionnaire from 397 Jordanian SME professionals across various managerial and accounting roles. The findings reveal that organizational culture exerts the strongest positive influence on CAIS adoption (β = 0.537, p < 0.001), followed by organizational structure (β = 0.296, p < 0.001) and organizational support (β = 0.226, p < 0.001). ITG also demonstrates a significant direct effect on CAIS adoption (β = 0.218, p < 0.001). Mediation analysis indicates that ITG significantly strengthens the relationship between organizational support and CAIS adoption (β = 0.089, p = 0.035), while its moderating effects on organizational culture and structure are not statistically significant. This research contributes to the digital accounting and Information Systems literature by providing nuanced insights into the conditional role of ITG in CAIS adoption, moving beyond traditional direct-effect models. From an empirical perspective, the findings recommend that policymakers and SME managers must priorities IT governance frameworks alongside organizational support mechanisms to facilitate effective CAIS implementation in emerging markets. Furthermore, prior studies have largely focused on technological determinants or direct effects of organizational factors, while providing limited insight into how IT governance shapes the relationship between organizational capabilities and CAIS adoption within SMEs in emerging economies.

Information doi: 10.3390/info17080754

Authors: Yang Meng Jinda Wang Tao Zhou Fei Niu Yanbo Wei

Benefiting from the merit of independent calibration without external reference, the shadow feature is investigated to retrieve wave steepness and wave height from X-band marine radar images. However, the wave period is currently required. Although the wave period could be achieved using the fundamental spectrum analysis technology from a radar image sequence, an external measuring device, like a wave buoy, is essential for the calibration of the wave spectrum. To solve this problem, an improved method for deriving significant wave height (SWH) is proposed by fusing the wave steepness extracted from the shadow feature of the radar image and the wavelength derived through rotated empirical orthogonal function analysis technology. Considering the attenuation relation of received echo in the distance direction, echo intensity calibration is adopted before extracting principal components. The collected X-band marine radar images are applied to verify the performance of the established SWH retrieval approach, with the observation of the wave buoy serving as the ground truth. Compared with the fundamental shadow statistical method, the correlation coefficient of the proposed approach rises by 0.08, while the root mean square error drops by 0.05 m.

Information doi: 10.3390/info17080753

Authors: Abdelouahed Moussaoui Nor-Eddine Azalmad Said Bahassine Khalid Housni

Part-of-Speech (POS) tagging is a foundational task in Natural Language Processing (NLP), yet it remains challenging for low-resource and morphologically rich languages such as Amazigh. This paper proposes a hybrid ensemble framework for Amazigh POS tagging that integrates three complementary models: a Bidirectional Long Short-Term Memory network (BiLSTM), a Conditional Random Field model (CRF), and a rule-based morphological analyzer (RBMA). Rather than treating prior results obtained on different corpora and tag inventories as directly comparable, the study evaluates all proposed components under a common 54-tag experimental setting based on the publicly available Amazigh Linguistic Dataset. Three ensemble strategies are examined: majority voting, validation-weighted voting, and logistic-regression stacking. An additional late-fusion ablation applies hard and soft RBMA constraints to CRF and Stacking outputs; hard masking degrades performance substantially, whereas soft masking is more robust but remains below unconstrained decoding. The best micro-level performance is obtained by the stacking ensemble, which reaches 98.51% Micro-F1/accuracy, whereas the boosting-like weighted ensemble obtains the strongest Macro-F1 among the ensemble variants, reaching 74.24%. These results show that hybrid ensemble methods can improve token-level accuracy in low-resource POS tagging, while also revealing a trade-off between frequent-tag accuracy and rare-tag robustness. The findings highlight the usefulness of combining neural, probabilistic, and rule-based information for Amazigh POS tagging, and point to class-balanced meta-learning and character/subword representations as important directions for improving rare and out-of-vocabulary categories.

Information doi: 10.3390/info17080752

Authors: Mehmet Akif Cifci Yousef Farhang Batuhan Öney Ziya Gökalp Ersan Fazlı Yıldırım Uğur Akbulut

Water resource management increasingly requires allocation methods that address rising demand, climate variability, and operational inefficiency. This study proposes an elite-transfer hybrid Genetic Algorithm–Particle Swarm Optimization framework for constrained water allocation under a weighted-sum scalarization model. The model combines four normalized objectives: minimizing water shortage, an operational cost coefficient, and allocation imbalance while maximizing utilization efficiency through an equivalent minimization term. A bidirectional elite-transfer mechanism links GA-based global exploration with PSO-based local refinement. The framework was evaluated using two capacity-constrained surrogate scenarios anchored to hydrometeorological records from Türkiye: Melekbahçe station (E21A033) in the Upper Euphrates Basin and Beşdeğirmen station (E12A003) in the Sakarya Basin. Daily streamflow records supported scenario construction, while precipitation and air-temperature data characterized local conditions. The proposed method was compared with standalone GA and standalone PSO, Differential Evolution, Grey Wolf Optimizer, a scalarized NSGA-II, adapted Kao–Zahara and Garg GA–PSO hybrids, and a no-elite ablation. All methods used the same objective formulation, normalization bounds, constraint-repair procedure, equal objective weights, tuning protocol, paired random seeds, and a budget of 10,000 objective-function evaluations. Performance was assessed through 30 paired runs per scenario. The proposed framework achieved the lowest mean scalar fitness values—0.1670 for Melekbahçe and 0.1752 for Beşdeğirmen—and reached the predefined convergence region after averages of 4587 and 4780 evaluations, respectively. All fitness and convergence improvements remained significant after Holm correction. Paired rank-biserial correlations ranged from 0.957 to 1.000 against the seven general comparators and were 0.824 and 0.781 against the no-elite ablation. The findings support the framework under the tested surrogate scenarios but do not establish Pareto-front dominance or universal superiority. Future work should examine measured operational data, additional basins, dynamic scheduling, alternative weights, and Pareto-based extensions.

Information doi: 10.3390/info17080751

Authors: Rajesh Kumar Narasimha Rao Vajjhala Ervin Ramollari Earta Joca

Automated weed classification supports site-specific weed management by enabling targeted interventions and reducing environmental and operational costs. This study presents a multi-seed evaluation of six deep learning configurations—YOLO26n-cls, Vision Transformer (ViT-B/16), DINOv2 (ViT-S/14) with linear probing and full fine-tuning, ResNet-50, and EfficientNet-B0—for nine-class weed classification using a stratified subset of the DeepWeeds benchmark. The models were evaluated using a leakage-checked train/validation/test split and repeated training across multiple random seeds to assess performance stability and statistical reliability. Among the evaluated approaches, DINOv2-FT and EfficientNet-B0 demonstrated the strongest overall performance, while statistical analysis showed that differences among several high-performing models were not consistently significant across seeds. The study also identifies the importance of appropriate transfer-learning strategies and hyperparameter selection, demonstrating that model performance can be strongly affected by optimisation choices. A reproducibility issue affecting ViT-B/16 was identified and transparently reported, leading to its exclusion from multi-seed statistical comparisons. An exploratory out-of-distribution probe using unlabelled Albanian field images further highlights the challenges of geographic domain shift and the need for locally collected, labelled datasets before reliable regional deployment. Overall, this work provides a systematic comparison of modern deep learning architectures for weed classification and emphasises reproducibility, statistical validation, and careful interpretation of model rankings.

Information doi: 10.3390/info17080750

Authors: Gary Reyes Roberto Tolozano-Benites Jorge Reyes Laura Lanzarini Waldo Hasperué Dayron Rumbaut Julio Barzola-Monteses Carlos George-Reyes

The sustained growth of urban areas has increased the complexity of managing services, infrastructure, and mobility, creating a need for advanced technological solutions capable of responding dynamically to rapidly changing environments. In this context, adaptive smart urban systems have emerged as an innovative alternative that integrates artificial intelligence (AI) to optimize real-time decision making. This study presents a systematic literature review and bibliometric analysis of 64 scientific articles focused on the frameworks underpinning these systems. The methodology applied is based on the selection and critical analysis of indexed scientific publications, enabling the identification of predominant approaches such as machine learning, deep learning, multi-agent systems, and reinforcement learning. The findings reveal a strong convergence between AI, the Internet of Things (IoT), and Big Data, as well as significant limitations in terms of interoperability, data governance, and scalability. It is concluded that, while the advances are promising, the consolidation of these systems requires a comprehensive approach that combines technological innovation, appropriate regulation, and social sustainability.

Information doi: 10.3390/info17080749

Authors: Ali Tariq Nagi Chiara Bellatreccia Andrea Borghesi Arianna Dondi Luca Pierantoni Daniele Zama Iria Neri Marcello Lanari Roberta Calegari

Synthetic data are increasingly proposed as a strategy for addressing data scarcity and representation imbalance in medical AI, particularly for paediatric populations and darker skin tones. However, visually plausible synthetic images may still contain clinically implausible features or fairness-relevant inconsistencies that are not adequately captured by automatic image-quality metrics. In this study, we present and empirically evaluate a clinician-guided framework for validating and selecting synthetic paediatric dermatology images. The framework combines a clinician-facing evaluation platform with structured assessments of visual realism, mask quality, diagnostic plausibility, confidence, and skin-tone relevance. Four clinicians with complementary expertise in paediatrics and dermatology completed 282 assessments of 93 real and synthetic images. Synthetic images were often rated as visually realistic but showed lower inter-rater agreement and weaker mask-quality assessments than real images. Clinician realism and confidence ratings were then used to divide 30 synthetic images into 18 approved and 12 non-approved images. To assess downstream utility, we compared a real-only ResNet50 classifier with classifiers augmented using all synthetic images, clinician-approved synthetic images, or non-approved synthetic images. Across three patient-level experimental splits, the clinician-approved condition achieved the strongest overall classification performance and the largest gains for the under-represented Dark-Skin subgroup. Because the Dark-Skin subgroup contained only seven patients and the synthetic subsets differed in size and disease composition, these fairness results should be interpreted as exploratory. The present study therefore provides evidence for clinician-guided validation and data curation rather than for a completed iterative generator-retraining process. Future work will evaluate whether clinician feedback can also support repeated generative-model refinement in larger, multi-centre datasets.

Information doi: 10.3390/info17080748

Authors: Firuz Kamalov Ahmed El Sayed Ikhlaas Gurrib Kweh Qian Long Ji Yeh Choi Ghassan Malkawi

Artificial intelligence applications in management, such as fraud detection and churn prediction, are frequently constrained by the class imbalance problem. Standard over-sampling methods, such as SMOTE, rely on local geometric interpolation, which assumes data convexity and struggles to model the disjoint structures typical of managerial datasets. We introduce Fourier-based Adaptive Spectral Synthesis (FASS), an over-sampling method that frames data generation as a signal reconstruction problem. By transforming the minority class data into the frequency domain via the empirical characteristic function, FASS isolates the global manifold structure from high-frequency sampling noise through an automated spectral filtering mechanism. We prove the L2-consistency of the underlying estimator. Empirical evaluations on credit and marketing datasets demonstrate that FASS favorably shifts the precision–recall trade-off compared to geometric baselines, reducing false positives and providing a theoretically consistent, parameter-free approach to learning from imbalanced data.

Information doi: 10.3390/info17080747

Authors: Xiaohui Du Yiwei Deng Xuelin Wang Biyang Ma Huifan Gao

News recommendation is a critical technology that helps users efficiently find content of interest from large candidate pools. Its core objective is to accurately model user reading interests. However, current news recommendation systems typically suffer from two key limitations: (i) they fail to suppress noise from a frequency-domain perspective, and (ii) they lack effective calibration for popularity bias within the embedding space. In this work, we propose a novel frequency-domain denoising and popularity-bias correction recommendation network (FDDP-RN) to address both challenges simultaneously. Our approach introduces spectral analysis into the news encoder. Specifically, we design a filtering mechanism that combines truncation and scaling to enhance high-frequency semantic components, improve text feature representation accuracy, and suppress redundant low-frequency components. In addition, we introduce a norm-scaling factor that dynamically calibrates the embedding distribution of cold-start news items, placing them on an equal footing with popular news items. This effectively improves the exposure of long-tail content without requiring extra user interactions. We conduct extensive experiments on three public datasets, namely, MIND-small, MIND-large, and Adressa. The quantitative results demonstrate that FDDP-RN achieves state-of-the-art performance. Notably, on the Adressa dataset, our model achieves an AUC of 75.36% and an nDCG@10 of 50.11%, outperforming the strongest baseline. Furthermore, cold-start fairness diagnostics on the MIND-small dataset reveal that our mechanism increases the top-10 long-tail exposure rate from 15.3% to 18.1% and reduces the exposure Gini coefficient from 0.991 to 0.987, indicating a better balance among recommendation accuracy, diversity, and fairness.

Information doi: 10.3390/info17080746

Authors: Zijing Cai Peixuan Li Jingwen Xu Shinan Li Jianan Song Jie Liu

The near-infrared (NIR) spectral responses under different tissue states are important for nondestructive soluble solids content (SSC) assessment of thick-peeled citrus, but the effects of peel removal and progressive tissue exposure remain unclear. Diffuse reflectance and transmittance spectra of Zigui navel oranges were collected at 650–1050 nm under four sequential states: intact fruit, first-slice fruit, second-slice fruit, and half fruit. Raw, first-derivative, and second-derivative spectra were analyzed using wavelength-wise paired tests, effect-size summaries, empirical response indices, and repeated validation of SSC prediction models. Slicing clearly altered the NIR spectral profiles, although these changes should not be interpreted as direct measurements of physical penetration depth. Diffuse reflectance spectra were more sensitive to surface-related changes, whereas transmittance spectra reflected cumulative optical-path responses. In 100 repeated stratified random splits, intact-fruit spectra showed only an average tendency toward more stable SSC prediction. Absolute prediction accuracy was weak for all tissue states, with the highest mean R2p being only 0.262 for intact fruit and negative mean R2p values for the other states. Thus, increased tissue exposure did not improve model generalization. This study mainly clarifies tissue-state-dependent spectral responses and provides a cautionary reference for SSC prediction in thick-peeled citrus.

Information doi: 10.3390/info17080745

Authors: Kamran Ibiyev

Turkic and Uralic languages share almost everything except ancestry: both are agglutinative, both stack long suffix chains onto a stem, both use vowel harmony, and both put the verb last. That coincidence is what makes the pairing worth testing. If a neural model learns where to place a stem–suffix boundary in one family, does the skill carry over to the other, given that nothing but structure is shared? We fine-tuned ByT5-small, a byte-level encoder–decoder, on segmentation pairs built from UniMorph 4.0 and Universal Dependencies for Azerbaijani, Turkish, Hungarian, and Finnish. To our knowledge, no earlier segmentation benchmark covers these four languages together, and the SIGMORPHON 2022 shared task included only Hungarian among them. These targets contain exactly one boundary at most, a stem plus one undivided suffix block, so the supervised task is single-boundary segmentation, not full morpheme decomposition. On this task, in-domain exact match is high across a 4×4 train-on-one, test-on-another matrix run over three seeds, 0.91 to 0.98, against a copy-input baseline of 0.02 to 0.20. Cross-family transfer is not: Finnish-to-Azerbaijani reaches 0.640 against 0.208 the other way (p<10−66), and an exploratory reading of the twelve pairs suggests surface typology tracks this gap more closely than genealogy does, though one high-overlap pair dominates and the sample is too small to settle the question. The central finding comes from a multi-boundary stress test: on 800 words we hand-annotated from Wikipedia (model-correctness agreement was perfect, Cohen’s κ=1.00), the models stall at two morphemes, and the 37% of words with three or more come out almost entirely wrong (one correct prediction in 897). Boundary-level scoring shows the shape of the failure: precision stays high, up to 0.97, while recall drops to 0.39–0.52, so the boundaries the models place are mostly right, they just place far too few. This ceiling is consistent with the two-part training targets rather than with the byte-level architecture, although only one architecture under one supervision regime was tested, so confirming the attribution will take retraining on fully decomposed targets. We release all code, data, and annotations to support that step.

Information doi: 10.3390/info17080744

Authors: Hamed Hilal AlYahmadi

Despite its ubiquity as a writerly practice in academic writing, there is little publicly available evidence that connects feedback-based revision practices with quantitative changes in research writing. This study proposes a framework based on artificial intelligence (AI) that aids in the assessment of revision-related development in student research exposés based on the Exposía academic writing and peer-feedback corpus (Ec). The raw files analyzed included 3247 records, with 16,068 feedback and comment records, 55 matched draft–final pairs of exposés, and 28 matched draft–final pairs of cases by score. For every text, features of NLP were computed, such as: word count, type–token ratio, moving-average type–token ratio over 100-word windows (MATTR-100), lexical density, readability, and adjacent-sentence cohesion. The differences between the draft and the final text were analyzed with paired statistical tests, and the human assessment scores were estimated using draft-Ridge Regression and draft-Random Forest models. Human assessment scores increased significantly from the draft (M = 5.26) to the final version (M = 7.55), t(27) = 9.04, p < 0.001, Cohen’s dz = 1.709. There was also a significant increase in word count, MATTR-100, and lexical density. Random Forest gave a better R2 (0.426) than Ridge Regression (R2 = 0.248). The exploratory feedback-improvement model resulted in a negative R2 due to a limited number of automatically merged feedback features, a lack of alignment between comment and revision, and a small number of matched scores. Therefore, AI-supported analysis identified writing development associated with revision and captured human judgment to some extent, but did not provide causal evidence of the independent contribution of peer feedback to writing improvement due to the observational design. The framework offers a repeatable benchmark and calls for enhanced detail in providing feedback–revision alignment.

Information doi: 10.3390/info17080743

Authors: Zhaobin Li Mingliang Mo Chenchong Du Zhanzhen Wei

Homomorphic encryption (HE) secure aggregation requires coordinate-aligned updates, whereas personalized federated learning (PFL) lets each client share a different parameter subset. Top-k sparsification widens this gap: per-client supports must be unioned and re-packed, exposing a data-dependent cleartext index. We propose Clustered-Support HE-PFL (CSS-FL), a two-phase protocol that clusters clients by their already-exposed Top-k support via k-means on binary indicator vectors. In an index phase, each cluster agrees on one shared sparse support by majority vote; in a value phase, clients encrypt exactly the agreed coordinates, which are aligned by construction and sum directly under CKKS, while non-shared coordinates remain local. Relative to full-union HE, the aggregation server gains no new information (clustering consumes only the routing indices Top-k already exposes), the only values ever decrypted are multi-client cluster means, and every party downstream of the server observes a coarse cluster-level index rather than per-client indices. On MNIST, Fashion-MNIST, and CIFAR-10 with K∈{2,4,8} clusters, LeNet (three seeds; five at the most heterogeneous setting) and ResNet-18 (three seeds), CSS-FL reduce CKKS encryption cost by 5.3–7.3× relative to a constructed Full-Union baseline at 0.25–0.5 the downstreamcleartext index footprint for K≤4 (the server-visible index footprint is unchanged), with personalized accuracy comparable to established PFL baselines (though below Ditto on CIFAR-10), and transmits 5–7× fewer end-to-end bytes including ciphertext expansion. An aggregation-width sensitivity study shows that averaging ≥4 updates, a width the protocol enforces via a minimum-cluster-size merge floor, collapses single-sample DLG reconstruction from 34.5 dB to ≈6 dB.

Information doi: 10.3390/info17080742

Authors: Kang Liu Jinfeng Zhang Hongxu Guan Chang Zheng Xinying Yang

With the rapid advancement of educational digitalization and generative artificial intelligence, knowledge graphs and AI agents have emerged as key technologies supporting smart higher education. To address challenges in current AI-assisted teaching, including insufficient support from structured knowledge systems, inadequate coordination across teaching processes, and delayed evaluation and feedback, this study proposes a collaborative framework integrating knowledge graphs and AI agents within an integrated teaching platform, and illustrates its integrated operational mechanism for knowledge organization, learning support, learning analytics, and teaching evaluation. Using the core course Nautical Navigation in the Navigation Technology Specialty at WHUT as a case study, the framework was implemented on the Chaoxing Smart Course Platform and applied to 459 students across two cohorts. The implementation achieved knowledge structuring and learning process visualization. The constructed course knowledge graph includes 336 knowledge points and more than 2700 associated learning resources and assessment items, while 30 instructional AI agents were developed and deployed to support different teaching and learning scenarios. The results indicate that the framework improves course knowledge organization, enhances student engagement and self-directed learning, enables visualization of learning processes and precision in teaching evaluation, and promotes a shift from experience-based to data-driven instructional decision-making. Compared with the previous cohort, students’ average daily learning time increased from 564 s to 618 s, participation rates in chapter quizzes, group discussions, and assignment completion all exceeded 90%, and more than 80% of respondents expressed willingness to continue using this learning model. The study provides a practical reference for AI-enabled teaching reform and digital transformation in higher education.

Information doi: 10.3390/info17080741

Authors: Dragoș-Ion Dogioiu Anca Andreea Morar Alin Dragoș Bogdan Moldoveanu Ana Magdalena Anghel Alexandru Ion Berceanu

Public speaking training and public speaking anxiety (PSA) treatment increasingly rely on technology-mediated systems, including web, mobile, desktop, and immersive virtual reality (VR) tools. This comprehensive review identifies and maps information and communication technology (ICT)-based public speaking interventions for educational training and anxiety-oriented treatment, focusing on platform choice, audience simulation, feedback timing, physiological and behavioral sensing, therapeutic framing, practitioner involvement, and evaluation methods. A PRISMA-inspired process documented identification and screening across Scopus, Web of Science, PubMed, IEEE Xplore, and ERIC for English-language journal articles and conference papers published between 1 February 2015 and 1 February 2025. From 6602 records, 82 participant-validated studies were included. No formal risk-of-bias assessment was conducted. VR accounted for 82.93% of interventions, while research prototypes comprised 72.0% of the evidence base, demonstrating the field’s strong emphasis on immersive and experimental development. At the same time, 64.6% of interventions provided no automated feedback and 36.6% used no physiological or behavioral sensing, highlighting opportunities for more responsive and data-informed systems. The findings can guide the development of future public-speaking training and treatment systems by highlighting recurring design components, promising implementation patterns, and priorities for stronger comparative validation.

Information doi: 10.3390/info17080740

Authors: Hangyu Yang Zhaoran He Lu Xu Rui Xue

In military confrontations, intentional suppression jamming can significantly degrade the reliability of friendly communication systems. By integrating conventional frequency hopping with spectrum sensing, cognitive frequency hopping technology can proactively avoid jammed frequency bands, thereby enhancing anti-jamming performance. This paper proposes a jamming detection and parameter estimation algorithm based on an improved YOLOv8 model. The proposed method extracts the time–frequency features of jamming signals and predicts their bounding boxes in time–frequency images. Based on the coordinates of the predicted bounding boxes, the center frequency, bandwidth, and temporal parameters of the jamming signals are estimated, thereby supporting spectrum sensing. Simulation results under MATLAB-generated signal conditions show that the proposed method achieves high detection accuracy and low mean squared relative error in the considered simulation scenarios. In addition, the proposed method maintains effective detection and parameter-estimation performance in composite jamming scenarios. This study provides a useful simulation-based reference for spectrum sensing in cognitive frequency hopping systems.

Information doi: 10.3390/info17080739

Authors: Mohammed El Assad Said Nouh Seddiq El Kasmi Alaoui Mohamed Azzouazi

Error-correcting codes improve the reliability of data exchanged between transmitters and receivers; achieving efficient and fast decoding is a major challenge. In this work, we propose an efficient, fast, and high-performing decoder named FastSyndISD. The first version of the proposed decoder uses hashing techniques to quickly find the information set that allows for error correction directly from the syndrome of the received word. Hashing information sets significantly reduce space complexity, instead of all correctable errors, compared to HSDec, LRDec, and EL-BoostDec. The second version utilizes a machine learning model (decision tree) to quickly find the information set that enables error correction directly from the syndrome of the received word. This second version does not require storing a hash table, since the information set is determined by a decision tree classifier model. The performance of the proposed decoder FastSyndISD is very encouraging regarding its bit error rate (BER). Indeed, it succeeds in correcting all errors whose weights are equal to or less than the correction capacity of the analyzed codes. Furthermore, a comparison with various other decoders highlights its effectiveness. A detailed study of its space complexity, its time complexity, and the factors influencing its performance has also been conducted, aiming to achieve an optimal BER while reducing computational time and memory usage. This comparison proves the significant success of the proposed decoder.

Information doi: 10.3390/info17080738

Authors: Dmaithan Almajali Robin Kabha Salwa Al Majali Samer Adnan Abdel-Hadi Lina Ashour Mohamed Fouda Ala Saleh Maria Tarawneh

Higher education is increasingly digitalized, resulting in increased reliance on e-learning systems, and thus, e-learning system success should be sufficiently evaluated. However, students worldwide have shown unproductive e-learning use. Hence, the e-learning systems success model after the outbreak needs revision. To validate the e-learning systems success, the role of monitoring quality was examined in this study using current e-learning and information systems success models. Data from 600 students were analyzed with structural equation modelling (SEM) run by SMARTPLS 4. Results showed that user satisfaction was positively impacted by information quality, system quality and service quality but not by monitoring quality. Results further showed positive impact of user system use on student satisfaction, subsequently impacting student loyalty. Also, results showed user satisfaction significantly impacting learning effectiveness. This study enhances the information systems literature through the inclusion of monitoring quality into the DeLone and McLean information systems success model and through examining its impact on user satisfaction, loyalty intention, and learning effectiveness in e-learning environments.

Information doi: 10.3390/info17080737

Authors: Mayank Kapadia Mohammad Masum

Accurate survival prediction in Liver Hepatocellular Carcinoma (LIHC) remains challenging because outcomes are shaped by clinical, molecular, and morphological heterogeneity. Although multimodal learning may improve prognosis, the contribution of each modality across clinically relevant survival horizons remains unclear, and many existing approaches provide limited interpretability. We present an interpretable multimodal framework for LIHC survival prediction using clinical variables, transcriptomic profiles, and histopathology whole-slide images from the TCGA-LIHC cohort. Survival prediction was formulated as horizon-specific classification at 1-, 3-, and 5-year endpoints. Histopathology patches were converted into morphology-focused textual descriptions using a vision–language captioning framework, aggregated into patient-level summaries, and embedded as predictive features. Caption quality was evaluated using automated semantic metrics and expert pathological review. Multimodal fusion was performed using a leakage-aware out-of-fold stacking strategy. The best fusion configurations achieved ROC-AUC values of 0.70, 0.74, and 0.66 for 1-, 3-, and 5-year prediction, respectively. Genomic features provided the strongest standalone signal at 1 and 3 years, while histopathology-derived captions contributed most clearly when combined with genomics at 3 years. These findings show that multimodal benefit is horizon-dependent and depends on modality complementarity rather than simply adding more data sources.

Information doi: 10.3390/info17080736

Authors: Weifeng Ren Ting Zheng Jie Zhang Yunzhong Chen Erkang Wei Chenxiao Liu Borui Fan Zhaiyuan Ji Yao Lu Jian He Yaxin Gao Shanqing Yu

Multi-expert large language model systems need to dynamically distribute user requests among multiple candidate experts to improve inference accuracy and system efficiency in complex task scenarios. To address the issues of existing routing methods relying on manual rules, unified discriminant boundaries, or large-scale annotated data, as well as high maintenance costs during expert expansion, this paper proposes an expert routing method, BGMM-LSE, based on positive sample semantic distribution modeling. The method constructs a positive sample set using only the requests historically successfully processed by each candidate expert, and maps the requests into dense semantic vectors through a pre-trained text feature extraction model; subsequently, a Bayesian Gaussian mixture model (BGMM) is independently trained for each expert to characterize the capability distribution of its successful requests in the semantic space. During online inference, the system encodes the target request into a feature vector, calculates the log-likelihood score combining the retained Gaussian component parameters and the smoothed covariance matrix of each expert, and uses Log-Sum-Exp to aggregate the probability contributions of different components. Finally, the request is routed to the expert model(s) with the highest score. Experiments were evaluated based on seven tasks: Math, GSM-Symbolic, HumanEval, Mbpp, MMLU, AIME1983–2025, and HellaSwag. The results show that this method achieves an average routing accuracy of 86.25%. The experimental results demonstrate that BGMM-LSE can provide stable request distribution capabilities for multi-expert large language model systems while maintaining interpretability and scalability.

Information doi: 10.3390/info17080735

Authors: Lei Bai Zhenguo Chen

Existing research on sentiment classification primarily focuses on textual contextual semantics while neglecting word co-occurrence and syntactic structure information. It also emphasizes semantic analysis of sequential structures but overlooks non-linear structural semantics. To address these limitations, we propose the DGCNCF (Dual-Graph Convolutional Network with Context Fusion) model for sentiment classification. The model employs a BiLSTM with an attention mechanism to extract textual contextual information. Subsequently, a text co-occurrence graph and a syntactic dependency graph are constructed separately. Meanwhile, contextual information is fused with node information within these graphs to obtain graph embedding representations of the text. Then, the model performs representation learning on the two graphs using GCN, thereby capturing the global word co-occurrence features and global syntactic dependency structure characteristics of the text. By integrating textual contextual information, word co-occurrence information, and syntactic dependency structure information, the model captures text features from semantic, lexical, and syntactic perspectives, thereby mitigating the limitations of relying on single-dimensional features. The validity of the model is verified on two public datasets, and the experimental results demonstrate that the model achieves effective sentiment classification performance.

Information doi: 10.3390/info17080734

Authors: Meijuan Huang Siyu Chen Bo Yang Xiaoyan Zhao

Keyword-based remote integrity auditing schemes effectively address the integrity of electronic medical records (EMRs) stored in the cloud. In practice, users expect to be able to perform flexible dynamic data updates while also protecting data privacy against a third-party auditor during the auditing process. However, existing schemes fail to simultaneously satisfy both requirements: they either incur prohibitive overhead for block-level updates or disclose to the auditor which EMRs match the target keyword and the number of such EMRs. To address this, we propose a new keyword-based auditing scheme for medical cloud. Specifically, we design a novel authentication identifier set. Unlike the keyword tags in Shen et al.’s scheme, this set aggregates the block hashes and thereby enables the auditor to verify integrity without obtaining sensitive information. Furthermore, we introduce a dynamic hash list. By updating this list during block insertion and deletion, the scheme eliminates the need to recompute the authenticators of subsequent blocks, significantly enhancing the efficiency of dynamic data updates. Security analysis confirms that the proposed scheme is secure. Performance analysis shows our scheme reduces block insertion and deletion overhead by over 60% compared to Gao et al.’s scheme, demonstrating high efficiency and practicality.

Information doi: 10.3390/info17080733

Authors: Shahil Raj Arun Pandian J

Grading short answers on a large scale is a significant challenge for academic institutions and online learning platforms. This study proposes a Square Root Exponential Inverse Document Frequency (SRE-IDF) weighting scheme to address this challenge. The proposed SRE-IDF scheme reshapes the decay curve of the traditional IDF. It suppresses common filler terms while preserving the discriminative importance of concept-bearing vocabulary for Automated Short Answer Grading (ASAG). The SRE-IDF is a parameter-free weighting scheme. It produces bounded term weights in the interval between e−1 and one. The proposed SRE-IDF was evaluated using three benchmark datasets: SciEntsBank, Mohler, and ASAG2024. The experimental results show that SRE-IDF consistently outperformed Traditional IDF, Smoothed IDF, and Best Matching 25 (BM25) across all three datasets. A comparison with a frozen Sentence Bidirectional Encoder Representations from Transformers (SBERT) baseline shows that SRE-IDF achieves competitive predictive performance while requiring substantially lower computational cost and offering greater interpretability. The analysis of statistical significance also reaffirmed the significance of the performance gains attained by the SRE-IDF. These findings support the claim that SRE-IDF is a robust and effective term-weighting strategy for similarity-based Automated Short Answer Grading (ASAG) in settings with heterogeneous and fine-grained evaluations.

Information doi: 10.3390/info17080732

Authors: Anna Tolia Stavros T. Ponis

Digital Twins are increasingly viewed as key enablers of real-time decision-making in supply chain management and logistics, yet the literature remains fragmented across application domains, decision problems, methodological approaches, and technological implementations. This systematic review examines how Digital Twins support real-time decision-making, understood not as a fixed response time threshold but as the temporal alignment between data refresh, decision generation, and system evolution. Following the PRISMA 2020 framework, a Scopus search conducted on 16 March 2026 identified studies in which Digital Twins were a central component, incorporated dynamically updated data, supported real-time or near-real-time decision-making, and addressed supply chain management or logistics problems. A total of 57 peer-reviewed studies were retained and synthesized using descriptive analysis, cross-tabulation, and thematic coding across decision problems, application contexts, solution methods, enabling technologies, implementation challenges, and future research directions. The findings show that real-time Digital Twin applications are concentrated mainly in production scheduling and planning, followed by routing and dispatching, resource allocation, disruption management, and inventory management. Methodologically, most studies adopt hybrid approaches combining simulation, optimization, and/or machine learning, reflecting the complexity of real-time operational decision-making. Enabling technologies were grouped into six functional layers, with data acquisition technologies receiving the greatest attention, while higher-level integration and enterprise system layers remain less developed. Reported challenges cluster around data, methodological, and systemic issues. The review concludes that real-time Digital Twins are emerging as integrated decision environments, but further research is needed on computational efficiency, interoperability, validation, and application in underexplored logistics domains.

Information doi: 10.3390/info17080731

Authors: Zhengda Chen Qizhi Wang Haoyang Li Jie Zhang Ben Hu Jie Liu

To address the evaluation requirements for agricultural robotic harvesting, this work presented a fruit-grasping assessment system based on tactile fusion analysis. Four piezoresistive pressure sensors were symmetrically integrated into the inner surfaces of a flexible gripper. A signal-conditioning circuit and a data acquisition module transmitted tactile signals to a Transformer–Mamba fusion network for feature extraction and target classification. After being trained on a dataset comprising 300 samples, the model extracted deep tactile features to distinguish among three target categories: citrus fruits, branches, and leaves. Classification outputs generated control commands for a robotic manipulator, enabling obstacle-avoidance retraction and precise harvesting operations. Experimental evaluations, conducted in both indoor and outdoor environments, demonstrated a target recognition accuracy of 93.12%. The manipulator response time was below 0.5 s, and the operational success rate was 90%. The proposed sensing system and algorithmic framework showed strong adaptability and supported quantitative assessment of grasping performance. The fruit detachment, compression damage, and plant-collision risks were effectively reduced while operational stability and harvesting efficiency improved.

Information doi: 10.3390/info17080730

Authors: Yuan He Haikun Lv Xing Li Lulu Zeng Lina Zhang Dengqi Yang Xiaowei Li

Software vulnerability mining is important for improving software reliability and security. Compared with sequence-based models, graph neural networks (GNNs) can explicitly model program structures such as syntax, control flow, data flow, call relations, and dependency paths, and have therefore been widely studied for vulnerability detection, localization, explanation, and repair. This paper presents a PRISMA-informed structured topical review of 87 studies and addresses five research questions concerning program graph representation, homogeneous and heterogeneous GNN architectures, Graph–LLM integration, evaluation reliability, and future research directions. The reviewed evidence shows that graph-based methods are most effective when vulnerability mechanisms can be faithfully represented through explicit structural relations. However, their reported performance remains strongly affected by duplicated samples, random function-level splits, noisy labels, incomplete repository context, graph-construction errors, and weak explanation protocols. Homogeneous GNNs provide efficient structural message passing but may mix different semantic relations, whereas heterogeneous GNNs preserve relation types more explicitly at the cost of greater graph-quality and computational requirements. Graph–LLM systems can improve semantic reasoning, repository-level analysis, explanation generation, and repair support, but their benefits should be evaluated together with memory consumption, inference latency, deployment complexity, and verification cost. This review further proposes minimum requirements for reliable vulnerability benchmarks and verifiable explanations, and develops a strategic agenda covering leakage-resistant datasets, uncertainty-aware graph construction, repository-level evaluation, cost-effective Graph–LLM collaboration, and graph-guided autonomous vulnerability repair.

Information doi: 10.3390/info17080729

Authors: Kabul Khudaybergenov Avazjon Marakhimov

Skeleton-based human action recognition is an important problem in applied vision systems, yet many existing approaches depend on a single skeleton descriptor or a single feature-learning mechanism. This restriction can weaken the representation of local body kinematics, long-range joint relations, and temporal dependencies within an action sequence. To address these limitations, this paper proposes HILF-GT (Hybrid Invariant Latent Feature Graph Transformer), a hybrid Graph Convolutional Network (GCN)-Transformer framework based on multiple spatio-temporal invariant latent features. The representation module constructs complementary structured tensors from skeleton graphs, inter-joint distances, adjacent-frame joint displacements, and inter-limb angles. Instead of transforming these descriptors into image-like maps for separate Convolutional Neural Network (CNN)-based classification, HILF-GT keeps their graph and temporal organization during learning. A local GCN branch models skeleton-aware kinematic patterns, whereas a graph-aware Transformer branch uses biased self-attention and cross-attention to capture dependencies among distant joints, frames, and latent-feature streams. A Perceiver-style latent bottleneck is further introduced to reduce the memory cost of global attention over frame-joint tokens. Experiments were conducted on four standard benchmark datasets, including NTU-RGB+D 60, NTU-RGB+D 120, NW-UCLA, and UTD-MHAD. The proposed method achieved 93.1% and 97.20% accuracy on the NTU-RGB+D 60 Cross-Subject and Cross-View protocols, 88.15% and 90.20% on the NTU-RGB+D 120 Cross-Subject and Cross-Setup protocols, 98.50% on NW-UCLA, and 97.50% on UTD-MHAD.

Information doi: 10.3390/info17080728

Authors: Ling Shang Zhizhong Liu Yuxuan Wu Xiaoyu Song Jian Yu Quan Z. Sheng

Currently, uncertain missing modalities pose new challenges to multimodal intent recognition. To tackle this issue, this work proposes an Attention Bidirectional Gated Fusion Based Multimodal Intent Recognition model under Uncertain Missing Modalities (named ABGFMIR). Firstly, ABGFMIR extracts the features of each modality (text, audio, visual) with the LSTM network, respectively. Secondly, ABGFMIR narrows the distances between audio, visual and text modality based on Central Moment Discrepancy (CMD), and then performs multimodal feature fusion through an attention bidirectional gated fusion method. Then, corresponding attention level prompts are generated based on the uncertain missing modalities situations of the current sample. The fused multimodal features are then input into the Transformer encoder and decoder, and the prompts are injected into the keys and values of the multihead self-attention layer to guide the Transformer to focus on the missing modes and dynamically adjust the attention distribution, enhancing the robustness of ABGFMIR to different missing modes. Finally, the features produced by the Transformer are fed into the classification layer for intent recognition. Simultaneously, the pre-trained model (AGNN) that trained with the complete modality is employed in the classification layer to guide the main module of ABGFMIR. Two public benchmark datasets (MIntRec and EMOTyDA) are adopted for performance verification. Compared with the other five baseline models, on the MIntRec dataset, ABGFMIR improved accuracy by an average of 2.68 and improved F1 values by an average of 3.24. On the EMOTyDA dataset, ABGFMIR improved accuracy by an average of 2.28 and improved F1 values by an average of 3.44.

Information doi: 10.3390/info17080727

Authors: Kalin Kopanov Tatiana Atanasova

Coordinated and bot-like activity on social media is usually studied with supervised detectors that need rich account data such as profiles, timelines, and follower networks, which is increasingly hard to obtain. We ask what can be established about a single account’s reply ecosystem from its publicly visible posts and replies alone, with no profiles, timelines, or follower data. In a case study of the reply ecosystem of an official political party account (23,953 replies by 1985 accounts, December 2025 to January 2026), we compute profile-free behavioral features covering text duplication, character-level entropy, timing regularity, reply latency, and post coverage, complemented by a co-commenting network analysis, and group active accounts with unsupervised density-based clustering. The clustering, combined with two transparent labeling rules, separates three behavioral tiers: templated amplifiers defined by text reuse, persistent responders with human-like text but extreme volume and coverage, and an organic remainder. The two non-organic tiers comprise 5.4% of accounts, yet produce 53.5% of all comments, a composition that is stable under resampling and threshold sensitivity analysis, with a failure mode that is only conservative, since over-strict settings leave a tier unassigned rather than reshaping it. The platform’s own spam flags, never used as input, rise steadily from organic accounts to templated amplifiers, consistent with the behavioral grouping.

Information doi: 10.3390/info17080726

Authors: Shweta Bansal Sumendra Yogarayan Siti Fatimah Abdul Razak

Indian mythological narratives contain long event chains, recurring characters, moral conflicts, interactions between human and divine agents, and source-specific narrative styles. General-purpose large language models can generate fluent text while losing character continuity, thematic relations, or source-supported events. This study presents MythoBiLLM, a parameter-efficient framework for summarization and continuation generation from Indian mythological texts. The framework combines a frozen Llama 3.2 3B-Instruct backbone, LoRA-based adaptation, and a gated BiLSTM narrative-memory adapter. A corpus of public-domain English translations from the Ramayana, Mahabharata, Bhagavad-Gita, Vishnupuranam, Harivamsha, Hindu Tales, and Indian Myth and Legend contains 3,684,838 word-level tokens and 6057 segmented passages. Evaluation covers language modeling, summarization, continuation generation, entity consistency, theme retention, component ablation, robustness, human assessment, and statistical testing. Relative to LLM+LoRA, the complete framework reduces average perplexity from 23.4 to 19.8. In controlled comparisons, the BiLSTM adapter achieves an MCS of 0.713 on both tasks, compared with 0.699 for the parameter-matched MLP adapter, 0.704 for independently trained long-context LoRA, and 0.708 for retrieval augmentation. Full MythoBiLLM reaches MCS values of 0.762 for summarization and 0.744 for continuation generation. After entity consistency and style alignment are excluded from MCS, the complete configuration retains the highest scores of 0.751 and 0.731. These findings support the complete framework on the evaluated corpus, while the controlled comparisons indicate a modest complementary contribution from the BiLSTM and do not identify it as the sole source of the performance gains.

Information doi: 10.3390/info17080725

Authors: Patricio M. Paccha-Angamarca Erwin J. Sacoto-Cabrera Víctor V. Velepucha-Bonett

The increasing convergence of enterprise architecture and IT governance frameworks, such as TOGAF 9.2, COBIT 5, and NIST CSF 1.1, with LegalTech regulations including GDPR 2016/679, eIDAS 910/2014, and NIS2 2022/2555, has created a growing need for rigorous and automated governance-validation mechanisms. However, to date, no formal approach exists to assess semantic conformance between a normative ontological model and its operational implementation in microservice-based systems, leaving critical governance gaps difficult to detect in legally sensitive environments. This paper proposes MALTG (Multidimensional Architecture for LegalTech Governance), a configurable and reusable formal framework that combines OWL 2 ontology engineering, a JSON-LD-based SDT (Structural Digital Twin), semantic conformance mapping, a hierarchical coverage function, and a conformance gap metric to support automated governance assessment and prioritised remediation. The framework accepts any organisational architecture as input, enabling application to arbitrary LegalTech case studies by replacing the reference SDT. The proposed framework models nine governance dimensions through an ontology of 59 classes and 15 properties and validates them against a semi-real, public-source SDT composed of 39 microservice components and 54 directed connections. The reference SDT was populated through an ontology-driven data-collectionprocess: a scraping campaign guided by the MALTG ontology (data/MALTG_Ontology.owl) harvested public information from the official portal of Ecuador’s Council of the Judiciary (Consejo de la Judicatura, CJ)—probing its technological-maturity level and digital-governance compliance—which was condensed into a single JSON-LD artefact, enforcing domain rigour and traceability on the search for LegalTech governance evidence. Experimental results demonstrated an overall ontological score of 82.6 and an SDT score of 73.7, with a mean conformance gap of 8.9. Five dimensions achieved full conformance, while the LegalTech dimension presented the largest gap. Graph-theoretic validation further confirmed monotonic improvement throughout the remediation sequence. Overall, the findings suggest that MALTG provides a formally grounded and reproducible approach for automating multi-framework LegalTech governance conformance assessment while maintaining semantic and structural traceability between normative models and operational architectures. As a proof of concept, this validation relies on a configurable, ontology-driven public-source (semi-real) SDT; external validity in real production LegalTech organisations remains untested and is left for future work.

Information doi: 10.3390/info17080724

Authors: Shan Lin Bensheng Yun Zhenyu Lin Jie Shen Qinghua Xu

Detecting small and distant objects in road traffic scenarios remains challenging owing to limited pixel resolution, cluttered backgrounds, and resource constraints on edge computing platforms. This work presents LCA-Net, a computationally efficient framework for small object detection that balances accuracy with model complexity. The framework incorporates three complementary designs: an Adaptive Deformable Downsampling Module (ADDM) that merges asymmetric and deformable convolution operations to improve spatial feature encoding while explicitly accounting for the parameter and computational cost of offset and modulation-mask prediction; a Cross-Scale Feature Fusion Pyramid (CSFFP) specifically engineered for minute objects, which augments multi-scale feature learning and enhances detection of far-field small targets; and a Lightweight Feature-Gated Detection Head (LFGDH) that employs channel–spatial attention to selectively emphasize informative features, thereby reducing both parameter count and computational cost. On Udacity, LCA-Net improves mAP@0.5 by 2.3 percentage points; on VisDrone2019, it improves mAP@0.5 by 1.7 percentage points. Across both benchmarks, the complete model reduces the parameter count by 25.58% and GFLOPs by 16.05% relative to YOLOv8-N. On the RTX A6000, LCA-Net-N reduces forward-pass latency from 1.82 to 1.63 ms, increases throughput from 549 to 613 FPS, and lowers peak GPU memory from 1180 to 1015 MiB. These results demonstrate a favorable accuracy–efficiency trade-off for real-time traffic perception.

Information doi: 10.3390/info17080723

Authors: Georgios Feretzakis Dimitrios Karapiperis Sarandis Mitropoulos

The proliferation of algorithmically generated malicious URLs presents a critical challenge for modern cybersecurity, requiring a shift from syntactic pattern-matching toward semantic understanding grounded in pre-trained transformer foundations. Building on our prior work on billion-scale semantic search and density-based campaign clustering, this paper presents a unified, deployable framework for real-time campaign discovery. The framework converts raw URL streams into dense Sentence-BERT embeddings and couples approximate nearest neighbor search with online density-based clustering, discovering emerging campaigns without prior knowledge of their number or shape. Our central finding is that the choice of semantic representation is decisive: a domain-focused embedding strategy yields near-perfect campaign separation, substantially outperforming full-URL representations. On live, in-the-wild threat feeds, the domain-focused representation recovers all 944 discovered campaigns at an Adjusted Rand Index of 0.990 and a mean campaign recall of 1.000, at 0.10 ms per URL. Under identical clustering, the full-URL representation reaches an Adjusted Rand Index of only 0.510 and recovers fewer than half of the campaigns. We add a SHAP-based explainability layer that attributes discovery decisions to interpretable structural patterns, and we expose the whole pipeline as an operational system with a REST interface for single- and batch-URL analysis. The encoder we employ is a discriminative representation model rather than a generative one; what we contribute is the representation-and-modeling discipline this setting demands, and the resulting guidance transfers to threat intelligence in an era of generatively produced attack campaigns.

Information doi: 10.3390/info17080722

Authors: Cláudia M. Viana

Historical document collections are increasingly available as digitised images and PDFs, but their conversion into reliable text remains affected by optical character recognition (OCR) errors, degraded pages, heterogeneous layouts, and domain-specific terminology. This study proposes a pathway-level framework for documenting and comparing conventional OCR, OCR followed by large language model (LLM)-assisted correction, and direct multimodal transcription. The framework is demonstrated using the Portuguese Agricultural and Forestry Surveys (1950–1958). A stratified validation sample of 45 pages was selected by visual quality, page type, and geographic coverage. Outputs were evaluated against manually verified reference transcriptions using content-normalised character error rate (CER) and word error rate (WER), document-condition analysis, paired statistical tests, and an entity-level semantic preservation assessment focused on place names, agricultural terms, and measurement expressions. Under the evaluated model and interface conditions, both LLM-based pathways produced lower mean CER and WER than the conventional OCR baseline, with the lowest values observed for direct multimodal transcription. Semantic preservation was also higher for the LLM-based pathways, although measurement expressions remained the most persistent risk, particularly in table-based pages. Downstream tasks were not directly evaluated. The findings support the framework as a method for validating text-extraction pathways before reuse, rather than establishing a universal ranking of tools.

Information doi: 10.3390/info17080721

Authors: Xudong Liu Yuedan Geng Qiang Ye

Online consumer reviews provide an important source of information for automobile purchase decisions, yet relatively little is known about the informational value of cumulative electronic word-of-mouth (eWOM) ratings that reflect long-term online reputation. This study examines whether cumulative eWOM ratings are associated with subsequent automobile sales and explores their relationship with subsequent changes in monthly ratings and rating entropy. We combine monthly automobile sales data from Chezhu Zhijia with more than one million individual consumer reviews collected from Autohome between January 2013 and December 2025. Review-level data are aggregated to the automobile model-month level to construct cumulative eWOM measures, and the hypotheses are examined using two-way fixed-effects models. The results show that cumulative eWOM ratings are positively associated with subsequent automobile sales. Higher cumulative ratings are also associated with more favorable changes in subsequent monthly ratings and lower rating entropy. The positive association between cumulative eWOM ratings and automobile sales becomes stronger as product age increases and varies with cumulative rating entropy. These findings remain qualitatively unchanged across a broad set of robustness and additional analyses. This study highlights the informational value of cumulative online reputation for high-involvement durable goods and provides new evidence on the conditions under which accumulated consumer evaluations are associated with market outcomes.

Information doi: 10.3390/info17080720

Authors: Olga Ussatova Vladislav Karyukin Yenlik Begimbayeva Galimkair Mutanov Zhanna Yessengaliyeva Yerlan Kistaubayev Medet Turdaliyev Daniyar Aitmukash Gulshat Baispay

The digital transformation of higher education has heightened the need for reliable mechanisms to verify the authenticity, integrity, and independence of academic documents. This study presents the design and practical implementation of a blockchain-based, multifunctional educational document verification system for the higher education environment of the Republic of Kazakhstan. The proposed system is presented by a hybrid multi-layered architecture that integrates a frontend web portal, a C# ASP.NET backend service, a PostgreSQL off-chain database, a Smart Bridge integration subsystem, and an Ethereum-compatible blockchain layer with Solidity smart contracts. The system separates operational educational data from trust-sensitive verification records: student, graduate, catalog, and document metadata are stored off-chain, while selected transcript and diploma-related data are registered on-chain and linked to transaction hashes. The Smart Bridge subsystem enables secure interaction with state information systems via SOAP/XML and digital signatures, improving the reliability of personal data used in academic record processing. The blockchain implementation uses transaction hashes, event logs, BlockScout monitoring, and role-based smart-contract authorization to support traceability and tamper-resistant verification. Experimental implementation showed an average gas consumption of approximately 226,125 gas per transaction. The proposed system demonstrates that blockchain extends existing centralized educational infrastructures by adding a transparent, auditable, and user-accessible trust layer for academic document management and verification.

Information doi: 10.3390/info17080719

Authors: Danah Algawiaz

The high rate of growth of misinformation on decentralized platforms causes a risk to public confidence and the integrity of decisions and requires a system of verification that is not only accurate but can be causally informed and interpreted via proxy causal metrics, reliable, and privacy-safe as well. The state-of-the-art federated learning (FL)-based fact-verification models mainly use correlation-driven patterns and do not provide ways to deal with causal reasoning, measuring formal reliability, or being resilient to Byzantine adversaries. This paper proposes a unified framework called CORE-FACT (causal optimization and reliability-enhanced fact-tracking) that can be used to conduct interpretable and robust misinformation detection in a distributed environment by combining causally informed optimization with reliability-weighted federated optimization. The proposed three-tier architecture includes: (1) a causal graph construction module, where variational attention is utilized to learn directed relationships of claims and evidence; (2) a reliability-weighted federated optimization module, where Byzantine-resilient aggregation (adaptive trust scoring) is achieved; and (3) an adaptive fact-tracking module, which is used to achieve fusion of multi-source evidence by combining hierarchical consistency verification with knowledge-graph embeddings. Empirical testing on the LIAR and FEVER datasets shows that CORE-FACT achieves 94.7% and 96.3% accuracy, respectively, outperforming state-of-the-art baselines by 3.5 to 4.8 percentage points in accuracy, with 17% lower latency than GEAR and 31% lower latency than DAGNN, and 98.2% robustness against 30% Byzantine attacks. Under differential-privacy guarantees verified at ε = 10.96, δ = 10−5 (Renyi DP composition, empirical membership inference validation committed for revision), CORE-FACT achieves a 31% reduction in false positives through explicit causally informed reasoning. These findings make CORE-FACT a scalable and interpretable framework that consolidates causal optimization, trustworthiness evaluation, and secure aggregation for next-generation federated misinformation detection.

Information doi: 10.3390/info17080718

Authors: Raj Bridgelall

Highway–rail grade crossing (HRGC) incidents in the United States declined substantially for several decades before stabilizing in recent years. Understanding this persistence is important because future safety improvements may depend on identifying locations where incident occurrence remains resistant to further reduction. This study developed an integrated framework to characterize persistent HRGC incident environments using 50 years (1976–2025) of Federal Railroad Administration incident records. Trend, structural-break, variance, and stationarity tests were first applied to determine whether the historical decline transitioned into a distinct persistence regime. A county-level persistence index (PI) was then developed to quantify the combined effects of incident burden and resistance to decline during the plateau period. Distributional analysis characterized the statistical behavior of the PI, while global and local Moran’s I statistics evaluated its spatial organization. Explainable machine learning methods were subsequently used to identify incident characteristics associated with elevated persistence. The results identified a statistically significant regime change around 2010. Prior to 2010, incidents exhibited a strong declining trend, whereas the subsequent period displayed a statistically significant but substantially weaker decline, lower variance, and behavior consistent with a persistence regime characterized by a markedly attenuated rate of improvement. The PI followed a strongly right-skewed distribution that was best represented by a bounded heavy-tailed unit log-logistic model, indicating that persistence is concentrated within a relatively small subset of counties. Spatial analysis revealed significant positive spatial autocorrelation (Moran’s I = 0.180, p = 0.001) and geographically coherent clusters concentrated primarily in the southeastern United States and several major freight-oriented regions. Explainable machine learning models identified train-operating characteristics, warning device contexts, movement patterns, and temporal conditions as key attributes associated with high-persistence counties. The findings demonstrate that the post-2010 incident plateau is sustained disproportionately by a limited number of geographically concentrated environments and provide a framework for supporting more targeted safety interventions.

Information doi: 10.3390/info17080717

Authors: Wael Zatar Hien Nghiem

An experimental morphology-based information extraction methodology is presented for locating reinforcing bars and estimating their effective ultrasonic scattering diameters in reinforced concrete (RC) members using ultrasonic pitch–catch (UPC) non-destructive testing combined with synthetic aperture focusing technique (SAFT) reconstruction. Reinforced concrete slab specimens with known reinforcement layouts were constructed and tested using a commercial ultrasonic device equipped with dry point contact shear wave transducers in a pitch–catch configuration. The collected ultrasonic data were post-processed using an in-house software package to reconstruct two-dimensional (2D) SAFT images of the specimens. In the reconstructed images, embedded rebars consistently produced characteristic bipolar scattering responses composed of dominant trough–crest waveform pairs. Based on these repeatable scattering response features, an empirical morphology-based procedure was developed in which rebar depth was estimated from the midpoint of the dominant trough–crest pair, while the effective ultrasonic scattering diameter was characterized using their vertical separation. The experimental results obtained from twelve reinforced concrete slab specimens demonstrated reliable depth estimation and a strong monotonic relationship between reconstructed scattering response width and nominal rebar diameter under the investigated testing conditions. Regression analysis indicated that larger rebars generally produced broader reconstructed bipolar scattering responses due to cylindrical wave interaction, diffraction, interference, finite transducer aperture effects, and SAFT reconstruction characteristics. The proposed methodology does not attempt to reconstruct the exact physical boundary of the reinforcement or solve the full elastodynamic inverse problem. Instead, the proposed methodology provides a practical and computationally efficient morphology-based information extraction framework that transforms reconstructed ultrasonic scattering responses into quantitative morphological descriptors for embedded reinforcement characterization. By extracting repeatable scattering response features from SAFT-reconstructed images and establishing an empirical mapping between these descriptors and reinforcement characteristics, the proposed approach enables the quantitative interpretation of ultrasonic images without requiring computationally intensive full waveform inversion.

Information doi: 10.3390/info17080716

Authors: Gowit Chanaken Osamu Uchida

Japan is one of the most earthquake-prone countries in the world. The 2011 Great East Japan Earthquake alone caused nearly 20,000 deaths and missing persons. International visitors and newly arrived foreign residents in Japan may face heightened risk during earthquakes, as limited prior experience with earthquakes and limited Japanese-language proficiency can restrict access to essential safety information. In addition, because disaster-preparedness education is often insufficient in countries with low disaster risk, they may be especially vulnerable in emergencies. This study designed, developed, and evaluated an augmented reality (AR) application intended to provide earthquake-preparedness and survival guidance in the Japanese context. The application was evaluated using a pre-test/post-test design with 40 participants from 10 countries, who completed a 14-item multiple-choice disaster-preparedness knowledge test and a user satisfaction survey. Participants’ knowledge scores improved significantly from the pre-test to the post-test (t(39) = 8.20, p < 0.001), with a large effect size (Cohen’s d = 1.30), and participants reported high satisfaction (4.55/5, SD = 0.61). These results suggest the potential of AR as an accessible educational approach for enhancing earthquake-preparedness knowledge among international visitors and newly arrived foreign residents in earthquake-prone regions.

Information doi: 10.3390/info17070715

Authors: Carlos Ayala-Tipan Lorena Recalde Edison Loza-Aguirre

Scientific production is expanding so quickly that research teams are struggling to track advances beyond their immediate specialization, especially in interdisciplinary areas where relevant work is scattered across venues and vocabularies. To reduce this overload at the group level, we propose an end-to-end pipeline that transforms structured bibliographic metadata into actionable topic recommendations for research teams. Starting from Scopus records, the method normalizes scholarly text, builds semantic author profiles using Sentence–BERT representations coupled with interpretable keyword descriptors, and forms candidate groups from co-authorship signals and profile similarity. For each group, the approach applies embedding-based topic modeling to generate candidate themes and ranks them using a relevance–novelty trade-off, enabling teams to surface directions that remain aligned with their collective agenda while still encouraging exploration beyond dominant or highly popular topics. Empirical evidence on a Scopus-derived corpus shows that embedding-aware descriptors support cleaner, semantically faithful representations than frequency-based baselines, strengthening downstream topic discovery and producing compact topic lists that are easier for teams to inspect, discuss, and adopt in collaborative planning.

Information doi: 10.3390/info17070714

Authors: Osama Al Maaini Khizar Hayat Khalil Al Ruqeishi Baptiste Magnier

The accuracy of spectro-temporal features for Makhaarij al-Huroof and Sifaat distinguishes between the ten canonical Qiraát recitation styles of the Holy Quran. However, real-world room reverberations blur formant contours and corrupt inter-word energies, thus making Qiraat discrimination difficult. The current dereverberation methods were designed to work under ordinary speech conditions and are not capable of preserving phonetic qualities for domain-specific purposes. This paper introduces a four-step, phase-consistent signal-processing approach prioritizing phonetic preservation over direct reverberation suppression. The four steps are: (1) adaptive noise-floor attenuation; (2) soft-voice activity detection using power-law boundary decay; (3) application-specific spectral contour adjustment from clean Quranic reference audio; and (4) Griffin–Lim algorithm-based phase correction. A total of 48 real-world room recordings were utilized for the evaluation of this approach based on Energy Ratio (ER), Spectral Contrast (SC), and Spectral Contour Stability (SCS)—measures specific to the Quran audio domain—alongside conventional speech-quality metrics. The proposed approach yielded the highest scores in three of seven metrics, namely SC (+40.11), SCS (+822.94), and PESQ (+1.251), alongside the second-highest Energy Ratio (+19.58 dB), while being superior to Spectral Subtraction, Wiener Filtering, and WPE Dereverberation approaches. Moreover, the perceptual enhancement was verified in a synthetic controlled experiment where the proposed approach scored an improved PESQ metric (+2.495; SNR −1.874 dB). The results illustrate the fact that an optimization for general-purpose metrics does not necessarily ensure phonetic preservation required for specific classification.

Information doi: 10.3390/info17070713

Authors: Cheng Jiang Jianyong Bi Huiqun Yu Mi Wen Lei Wu Rolf Findeisen

Cyber-physical power systems (CPPSs) are a crucial component of smart grids, integrating physical power systems with advanced information and communication technologies to achieve efficient, reliable, and intelligent energy management and control. However, with the widespread deployment of information technology, CPPSs face increasingly severe security threats and privacy protection challenges, such as data leakage, identity forgery, and impersonation, which can compromise the secure and stable operation of CPPSs. To counter these threats and protect privacy, cryptographic technique is developed to provide fundamental and powerful tools, supporting secure authentication, data integrity checking, privacy preservation, and trusted communication between connected devices in CPPSs. We systematically review the research progress on security authentication and privacy protection in CPPSs from a cryptographic perspective. Our survey analyzes the major security threats faced by CPPSs, along with the impact of various attacks on system data. We explore mainstream cryptographic algorithms, including digital signatures, key agreement protocols, signcryption authentication, homomorphic encryption, and blockchain-based security mechanisms that are capable of resisting cyber attacks and ensuring reliable decision-making and control in CPPSs. This work also provides the trends and challenges regarding the intersection of cryptography and networked control, blockchain scalability, and convergence of cryptography and artificial intelligence in CPPSs.

Information doi: 10.3390/info17070712

Authors: Saahoon Hong Hunhui Na

Background/Objectives: This scoping review examined recent evidence on the use of transformer-based language models, encompassing encoder-only architectures (e.g., BERT and its clinical variants) and generative large language models (LLMs; e.g., GPT-4 and Llama), to support clinical decision making from unstructured clinical notes, with implications for behavioral-health services where narrative documentation is central. Methods: Following PRISMA-ScR guidelines, PubMed, PsycINFO, and Web of Science were searched for peer-reviewed studies published between 1 January 2023, and 5 August 2025. Studies applying transformer-based language models to clinical narratives for healthcare tasks and reporting evaluative outcomes were included. We extracted data on clinical tasks, model architectures, enhancement strategies, and evaluation metrics; mapped each study by primary purpose, care setting, and primary model approach; and charted reported validation design, direct human comparison, fairness assessment, workflow evaluation, and clinical deployment. Results: Thirty-six studies were included. Information extraction/de-identification and classification/prediction predominated, whereas summarization/generation was less commonly represented. Model approaches appeared to align with task characteristics: encoder-only and decoder-only systems were frequently used for extraction, encoder–decoder systems for generation, and hybrid or pipeline-based approaches for classification and prediction. Standard task-specific metrics (e.g., F1 and AUROC) predominated, whereas evidence beyond retrospective task performance, including direct human comparison, fairness assessment, workflow evaluation, clinical deployment, and temporal or external validation, was rare. No included study evaluated a transformer-based language model application within a behavioral-health service or behavioral-health workflow. Conclusions: Transformer-based language models have been applied across diverse clinical-note tasks, but the evidence base more strongly supports retrospective task feasibility than transportability, equitable performance, workflow benefit, or safe clinical deployment. Future research should prioritize transparent reference standards, external and prospective validation, clinically meaningful human comparison, and equity-focused evaluation, including direct studies in behavioral-health services.

Information doi: 10.3390/info17070711

Authors: Pengfei Lyu Wei Xie Toshio Eisaka

Sustaining learner motivation remains a persistent challenge in higher education Japanese-language e-learning, where learners often study with limited social presence and personalized encouragement. This paper proposes a preliminary Design Framework for Motivational Robots in E-Learning (DFMRE), derived through a retrospective, theory-guided synthesis of two previously published empirical studies on robot-assisted Japanese-language learning. The synthesis interprets the prior findings through Self-Determination Theory and self-efficacy theory and formulates five candidate design principles: human-affine compact hardware, multi-level learner-selectable gestures, calibrated vocal encouragement, learner-initiated interaction protocol, and content-independent system integration. To provide an initial external check, nine domain experts with diverse backgrounds in education, educational technology, human–robot interaction, and related fields provided a preliminary appraisal of the principles in terms of clarity, feasibility, transferability, and overall usefulness. The framework received broadly favorable ratings, including a mean overall usefulness score of 4.33 on a 5-point scale, while expert comments highlighted the need for clearer operational definitions and flexible interaction modes. Because the empirical base consists of two small-sample Japanese-language learning studies conducted at one institution using one robot platform, DFMRE should be read as an early, context-grounded, falsifiable proposal rather than as a confirmed model for higher education e-learning in general.

Information doi: 10.3390/info17070710

Authors: Tianyue Wei Xin Lai Yidan Liang Chengwei Zhang Rui Kang

Dynamic aeronautical information is still widely disseminated through textual Notice to Air Missions (NOTAMs), while message-oriented storage cannot directly maintain the evolving states of affected objects. The objective of this study is to determine whether NOTAM-derived object events can be organized into traceable bitemporal states that support accurate and efficient current and historical access. To this end, this paper proposes BiTE, a bitemporal event-centered database framework that connects object-level event evidence, historical state versions, and materialized current-state projections. By integrating business and system time with NOTAM-specific lifecycle rules, BiTE supports state maintenance, historical reconstruction, and source traceability. A MongoDB-based prototype was evaluated using 44,591 NOTAMs from five major U.S. aerodromes. Independent manual validation showed 96.14–100% agreement across object identification and lifecycle-maintenance tasks. Across 1500 manually verified queries, BiTE achieved F1 scores of 99.43% and 98.66% for current-state and airport-overview retrieval, respectively, and a historical hit rate of 95.80%, outperforming representative message-oriented, relational-bitemporal, and RDF-based implementations. Mean query latency remained below 3.7 ms, while functionally equivalent ablations confirmed the performance contribution of the layered architecture. These results demonstrate that BiTE enables accurate, traceable, and efficient object-state management for dynamic aeronautical information.

Information doi: 10.3390/info17070709

Authors: Sepideh Kalateh Nastaran Farhadighalati Sanaz Nikghadam-Hojjati Jose Barata

Research on creativity often treats emotions using broad categories such as positive versus negative affect, which obscure the role of specific emotional states in creative performance. This exploratory study examines how fine-grained emotions relate to linguistic creativity in improvised speech. Thirty adult participants completed speech tasks across happy, neutral, and sad conditions, producing 90 speech samples. The responses were transcribed and evaluated using TTCT-inspired creativity dimensions: fluency, flexibility, originality, and elaboration. Fine-grained emotional states were estimated from the transcripts using automatic emotion recognition, and their associations with creativity scores were examined through statistical analysis. The findings suggest that linguistic creativity in improvised speech can be better understood by considering specific emotional profiles rather than relying only on broad positive or negative affect categories. The study contributes an exploratory methodological framework for future research on affect-aware creativity support systems and Computational models of emotion–creativity interaction.

Information doi: 10.3390/info17070708

Authors: Bingling Huang Jiahao Li Ziyu Li Hao Jiang Mingxiang Yang Xiaoxia Li Xiaohui Niu

Alzheimer’s disease (AD) is a severe neurodegenerative disorder whose pathological progression is closely associated with the reduced binding affinity of apolipoprotein E ε4 (ApoE4) for amyloid-β (Aβ), which impairs Aβ clearance. Existing computational molecular design approaches are largely limited to single-target optimization and therefore lack the capacity for synergistic dual-target modulation. Herein, we proposed EvoPlay-MuZero, a hybrid computational framework incorporating a dual-peptide bridging (DPB) strategy. The framework adopted latent-state planning in MuZero reinforcement learning to enhance exploration and sequence-generation efficiency in high-dimensional sequence spaces. For the first time, it enabled the automated design of bispecific peptides targeting ApoE4 and Aβ, thereby forming a synergistic molecular bridge via a flexible linker. A full-process pipeline for structural and energetic evaluation was established by integrating AlphaFold3 and PDBePISA. Benchmark experiments on the 1SSC, 2CNZ, and 3R7G datasets demonstrated that EvoPlay-MuZero substantially outperformed the vanilla EvoPlay in convergence speed and the yield of valid generated sequences. Specifically, the optimal DPB molecule (15 × 25-3A) achieved a calculated interfacial solvation energy score (ΔiG) of −29.5 kcal/mol, demonstrating substantially enhanced interface-stabilization properties compared to the native baseline control. This study provides a novel molecular intervention strategy for adjuvant therapy in AD and highlights the considerable potential of reinforcement learning in multi-target drug design.

Information doi: 10.3390/info17070707

Authors: Xiao Zhou Wenbing Liu Jun Wang Yilong Han

To address the problem that hotel recommendations in tourism activities do not consider the spatial relationship between hotels and scenic spots and the cost of tour routes, we construct a tourism hotel recommendation model based on ISTING-AGNES machine learning and an IDFST optimal route algorithm. Firstly, a scenic spot spatial clustering model based on the ISTING-AGNES machine learning algorithm is constructed, including a scenic spot ISG spatial topological model based on the neighborhood cell growth algorithm and an ISTING-AGNES machine learning algorithm based on the scenic spot ISG spatial topological model, which can realize spatial dimension reduction in tourist cities and construct tourism sub-regions for recommending scenic spots and hotels. Secondly, taking the tourism sub-regions as the modeling scope, a tourism hotel recommendation model based on the IDFST optimal route algorithm is constructed in which a closeness model between the tourists’ interests and the attributes of scenic spots in the sub-region is established to recommend the most matched scenic spots for tourists. Then, based on the recommended scenic spots, a tourism sub-interval optimal route algorithm based on IDFST and a tourism hotel recommendation model based on the optimal route decision forest algorithm are established to search for the global optimal tour route—with hotels as the starting and ending points and scenic spots as nodes—and to recommend the hotel with the most cost-effective tour route for tourists. The experiments prove that the constructed algorithm can output the hotel with the best geospatial location and the lowest tour route cost. Under the experimental conditions, compared with the hotel recommended by the weighted centroid positioning algorithm, the cost optimization rate reaches 5.98%. Compared with the greedy mountain climbing algorithm and the greedy BFS algorithm, the cost optimization rates reach 14.73% and 16.03%, proving that the constructed algorithm is feasible and advantageous over the traditional algorithms.

Information doi: 10.3390/info17070706

Authors: Marijela Miličević Mia Rovis Ratomir Karlović Sandi Baressi Šegota Vedran Mrzljak Ivan Lorencin Darko Etinger

Large language models (LLMs) are increasingly integrated into educational settings, yet their pedagogical reliability remains insufficiently understood. Beyond overt hallucinations, which informed users readily recognize, a subtler failure mode consists of explanations that are coherent, authoritative, and pedagogically plausible while harbouring hidden conceptual flaws, responses we term Socratic traps. This paper introduces SocraticTrap-CS, a publicly available benchmark that probes the capacity of open-weight LLMs to generate such strategic misconceptions on demand. A single structured prompt explicitly elicited three outputs per concept (a correct explanation, an overt hallucination, and a strategic misconception), yielding 735 expert-annotated response segments from seven open-weight models across 35 core concepts in algorithms and data structures, programming languages and paradigms, databases, computer networks, and operating systems. Three domain experts independently annotated each segment using a three-class schema, achieving near-perfect agreement (Fleiss’ κ=0.9487). Because models were explicitly instructed to produce the misconception, the central metric quantifies adversarial instruction-following capacity rather than the base rate of such errors in naturalistic use and should be read as a conservative upper bound on model capability. Under these conditions, compliance reached 91.7% overall (100% for three models; 57.1% for the smallest model, Mistral 7B, whose lower rate plausibly reflects weaker instruction-following rather than greater safety). Expert-judged persuasiveness was moderate to high, errors were predominantly conceptual rather than factual, models differed significantly, and no statistically significant domain-level differences were detected. The benchmark reframes the evaluation of educational LLMs around pedagogical trustworthiness rather than factual correctness alone.

Information doi: 10.3390/info17070705

Authors: Jingyang Wang Jie Ma Jiaxi Zhang Zehan Li Min Huang

To address the problems that existing UAV RF signal azimuth estimation methods rely on idealized simulation data and lack accuracy and robustness in complex environments, a high-precision azimuth estimation method based on an improved ResNet, namely the Dual-Branch Circular-Regression Network, is proposed. Firstly, UCA-8 array data is generated from measured single-channel RF signals, and non-ideal factors such as channel mismatch and mutual coupling among array elements are incorporated to simulate the real RF receiving environment. Secondly, ResNet is improved from three aspects: input normalization, dynamic dual-branch (DDB) learning features and periodic angle regression. The input normalization strategy based on Per-Sample Complex Root Mean Square (PSCRMS) is adopted to improve the adaptability of the model to signal scale changes. The DDB structure is adopted to adaptively fuse I/Q spatiotemporal features with a spatial covariance statistical prior to enhance the spatial feature expression ability in complex scenes. A periodic angle regression method based on Unit Circular Vector Representation (UCVR) and the Huber Loss (GAH Loss) of geodesic angle distance is adopted to realize periodic angle continuous modeling and suppress abnormal angle errors. Finally, comparative and ablation experiments are conducted on the constructed UCA-8 dataset. The experimental results show that compared with the baseline ResNet, the Dual-Branch Circular-Regression Network achieves 85.8%, 95.3%, and 81.5% reductions in MAE, RMSE and P95, respectively, and maintains higher estimation accuracy and good robustness under low signal-to-noise ratio, hardware mismatch and co-frequency dual-source interference.

Information doi: 10.3390/info17070704

Authors: Yanping Chen Junjie Zhang Chi Zhang

This work addresses coverage formation control for the Multi-Rover System (MRS) in extraplanetary environments such as Mars or the Moon, where Global Navigation Satellite System (GNSS) signals are unavailable and uncertain self-localization significantly degrades coverage performance. Existing coverage strategies predominantly assume perfect localization, which is unrealistic for GNSS-denied planetary surfaces. This paper presents a novel multi-rover coverage formation control algorithm that combines guaranteed Voronoi partitioning with belief space planning. The core contributions are (i) a guaranteed Voronoi partitioning framework that provides deterministic bounds on true coverage cells under localization uncertainty; (ii) a dual-layer optimization architecture integrating Extended Kalman Filter-based self-localization with centroid-error minimization; and (iii) a belief space planning approach that predicts system state evolution and infers optimal control inputs over a receding horizon. Simulation results under multiple density distributions and rover configurations demonstrate faster convergence and lower coverage cost compared to conventional Lloyd-based methods. By extending existing coverage strategies, our approach supports rapid and adaptive deployment of multi-rover networks in GNSS-limited environments, providing a promising solution for large-scale planetary exploration.

Information doi: 10.3390/info17070703

Authors: Yanzhi Liu Ming Meng

Accurate differentiation of Alzheimer’s disease (AD) and frontotemporal dementia (FTD) is clinically important because their management strategies differ. This study aimed to develop and evaluate a Dynamic Threshold Graph Convolutional Network (DT-GCN) and to systematically compare four EEG-based functional connectivity (FC) measures—Pearson correlation, phase-locking value (PLV), Granger causality, and copula analysis—for classifying AD, FTD, and healthy controls (HCs). In contrast to fixed graph binarization, DT-GCN updates the connectivity threshold at each training epoch according to training-fold loss. The model is trained under a multi-task objective combining classification, reconstruction, and contrastive losses, with the loss weights adjusted across three stages of training. Resting-state 19-channel EEG recordings from a public dataset (DS004504) comprising 36 AD, 23 FTD, and 29 HC participants were segmented into non-overlapping 8 s epochs. FC estimates were aggregated to obtain a subject-level adjacency matrix for each measure, with one-hot node identity and node degree as node features. Classification was performed at the subject level using stratified five-fold cross-validation with normalization and hyperparameter selection confined to training folds. Under the broadband setting, copula-based FC with DT-GCN yielded the highest observed three-class accuracy of 0.82±0.07 (macro-F1 0.79±0.07), compared with 0.45±0.06 accuracy and 0.36±0.06 macro-F1 for the standard GCN baseline. However, the small single-center cohort (N=88, including 23 participants with FTD) and the resulting small test folds limit the precision of the performance estimates and preclude robust inferential comparisons among FC methods and classification scenarios. These exploratory, dataset-specific findings require independent external validation and should not be interpreted as generalizable clinical performance or validated clinical biomarkers.

Information doi: 10.3390/info17070702

Authors: Ratchaneekorn Khamphukun Warawut Narkbunnum

Machine learning is increasingly used in entrepreneurship analytics to predict startup outcomes, frequently reporting accuracy above 0.90, yet whether such performance reflects a genuine ex-ante signal or methodological artifact remains unclear and consequential for investors, accelerators, and innovation-policy agencies. This study evaluates startup-outcome classification under a leakage-controlled, calibration-first protocol using two Crunchbase-derived datasets (66,368 firms; a 923-firm engineered-feature set), three success constructs, and three model families under five-fold stratified cross-validation. Removing outcome-correlated, survivorship-accumulating features lowers the area under the receiver operating characteristic curve by 0.05 to 0.09 on the large dataset, with every paired 95% confidence interval excluding zero, and by 0.19 on the engineered dataset; an independent study on the same 923-firm data without leakage control reports 88.1% accuracy. The leakage-controlled performance level is modest (0.66 to 0.77). Calibration rankings diverge from discrimination rankings: gradient boosting is well calibrated (expected calibration error of 0.006 to 0.024), whereas logistic regression shows large calibration error on imbalanced constructs (largely an artifact of class weighting rather than an intrinsic model property); post hoc isotonic recalibration then removes most of the error. The contribution is a reusable evaluation protocol for entrepreneurship analytics. Findings are associational and specific to the analyzed samples.

Information doi: 10.3390/info17070701

Authors: Liping Wang Yaozheng Zhao Yan Chen Guangyong Xi

To address the issues of rolling bearing fault vibration signals being susceptible to noise interference and the support vector machine (SVM) relying on manual parameter settings, this paper proposes a fault diagnosis method based on ICEEMDAN and AIHO−SVM. Firstly, the Hippopotamus Optimization Algorithm is improved by incorporating Chebyshev chaotic mapping, refraction opposite learning, dynamic weighting, adaptive step size, and guided learning strategies, thereby enhancing convergence accuracy and speed. Secondly, ICEEMDAN is employed to decompose vibration signals for noise reduction, and the effective intrinsic mode function (IMF) components are selected according to the mutual information criterion to reconstruct the signal. Nine time−domain statistical features are then extracted to construct the fault feature vector. Thirdly, AIHO is used to collaboratively optimize the penalty factor and kernel parameters of SVM, establishing an AIHO−SVM classification model. Finally, the proposed method is validated on bearing datasets from Case Western Reserve University and Huazhong University of Science and Technology. Experimental results show that the average diagnostic accuracies on the two datasets reach 99.07% and 99.09%, respectively, demonstrating the effectiveness of the proposed method for rolling bearing fault diagnosis.

Information doi: 10.3390/info17070700

Authors: Fahim Sufi Fatematuz Zohra

Despite advances in computational social science, online conflict discourse is still commonly analyzed through sentiment, stance, or topic models, leaving limited insight into the social and psychological mechanisms embedded in digital war narratives. This study develops a theory-informed probabilistic framework for detecting discourse-level indicators of Social Identity Theory, Moral Foundations Theory, Threat Appraisal, Cognitive Distortion, and Deindividuation in Russia–Ukraine war discourse. The empirical design uses 48,201 tweets in total: 10,815 tweets collected between 1 January and 28 June 2022 for model development and primary analysis, as well as and an external validation corpus of 37,386 Russia–Ukraine cyberwar-related tweets—collected from 30,706 users across 54 languages between October 2022 and April 2023—for temporal robustness assessment. The primary corpus contained 10,815 unique tweet identifiers, 10,229 unique textual records, 586 repeated textual items, a textual uniqueness rate of 94.58%, 6646 English tweets (61.45%), and 32,260 retweet engagements. Methodologically, the framework combines contextual language representations, theory-aligned linguistic cues, temporal signals, engagement features, and graph-based indicators. These signals are used to infer latent constructs and are evaluated through calibration, ablation testing, human validation, and cascade comparison. Empirically, Deindividuation was the dominant construct (1654 posts, 15.29%), followed by Cognitive Distortion (525, 4.85%) and Threat Appraisal (503, 4.65%). Co-activation analysis showed the strongest overlap between Deindividuation and Cognitive Distortion (Jaccard = 0.26). Validation diagnostics indicated internal lexical consistency (r=0.88 for Deindividuation), 93% rumor calibration, 91% bootstrap stability, and improved baseline performance (F1 = 0.72; Brier = 0.12; cascade log-likelihood = −865). The findings demonstrate that theoretically grounded probabilistic modeling can provide scalable, interpretable, and temporally validated insight into psychological patterns in digital conflict discourse.

Information doi: 10.3390/info17070699

Authors: Guochen Zhang Qing Ye Xiaobo Li Zhe Song

Wind power forecasting plays a crucial role in power dispatch and safety management of wind farms. However, the insufficient integration of multimodal heterogeneous data and the limitations of conventional graph construction strategies significantly restrict forecasting performance. Existing approaches either rely on simple feature aggregation, which cannot fully capture cross-modal dependencies, or adopt predefined or single-criterion graph construction methods that fail to characterize complex turbine relationships involving spatial, temporal, and nonlinear correlations. To address these challenges, this paper proposes a Multimodal Adaptive Fusion Graph Neural Network (MAF-GNN) for short-term wind power forecasting. First, a Modality-Aware Representation Learning (MARL) module is developed to extract informative multimodal representations by modeling modality-specific characteristics and cross-modal dependencies through attention-based fusion. Second, an Adaptive Graph Learning with Multi-Similarity (AGL-MS) module is introduced to parametrically integrate four complementary similarity priors—geographic distance, Dynamic Time Warping (DTW), Maximal Information Coefficient (MIC), and cosine similarity—for adaptive turbine correlation graph construction. Furthermore, a Credibility-Modulated Graph Convolutional Network (CM-GCN) is developed to reduce the influence of unreliable node information during message propagation. Extensive experiments conducted on the SDWPF dataset demonstrate that MAF-GNN reduces MAE by 14.0–21.3% compared with sequential baselines and achieves 5.3–10.2% improvement over spatiotemporal graph-based models. Ablation studies further verify the complementary effectiveness of each proposed module in improving forecasting performance.

Information doi: 10.3390/info17070698

Authors: Claudiu Coman Costel Marian Dalban Vlad Bătrânu-Pințea Georgiana Aron Lucian Marina

Fake news detection has become a major research topic at the intersection of artificial intelligence, data mining, and information security. In this paper, we evaluate the performance of English-trained algorithms on English translations of Romanian-sourced news articles, using a translation-mediated cross-domain evaluation design. The study is based on source code generated with the assistance of artificial intelligence systems for a set of machine learning and transformer-based models. The code was subsequently implemented in Google Colab. 2026, trained on international benchmark datasets, and tested on Romanian news content. This design allowed the rapid prototyping of multiple detection pipelines and the systematic observation of their behavior in a media environment different from that represented in the training corpora. The models were evaluated comparatively using standard classification metrics, including accuracy, precision, recall, and F1-score, complemented by additional indicators relevant to model robustness and practical usability. The experimental results revealed significant differences in performance across algorithms when applied to English translations of Romanian-language news content after training on international datasets. However, this study does not provide a direct comparison between model performance on the international benchmark datasets and the Romanian test corpus; therefore, the gap between the international training corpus and the Romanian-sourced test corpus is interpreted as an exploratory limitation and as a direction for future research. Based on these findings, we propose an empirical classification of the tested models according to their predictive effectiveness, their contextual robustness across linguistic environments, and their operational relevance as filtering tools for institutional monitoring. The results show that AI-assisted coding workflows can provide a viable starting point for reproducible misinformation research, but they also underline the limitations of directly transferring models trained on non-Romanian data to local media ecosystems. The study offers both a replicable evaluation framework and practical insights for institutions involved in strategic communication, public security, and the monitoring of information threats.

Information doi: 10.3390/info17070697

Authors: Boris Astudillo Marco Santórum Jose Aguilar Mayra Carrión-Toro Patricia Acosta-Vargas

Data Analytics methodologies provide structured approaches for transforming organizational data into actionable knowledge. However, many existing methodologies emphasize technical activities while offering limited support for stakeholder participation, user-centered validation, and organizational adoption. This study presents the construction of MIDA5, a Data Analytics methodology for Business Process Improvement developed using the Design Science Research paradigm. The research combined a comparative analysis of existing Data Analytics methodologies with an empirical experimentation process conducted in an organizational environment. The experimentation involved the execution and analytical deconstruction of a previously implemented Data Analytics methodology to identify operational limitations, stakeholder-related challenges, and methodological gaps. The findings were synthesized into design requirements, methodological components, and design needs that guided the construction of MIDA5. The resulting artifact incorporates principles of Business Process Analytics, User-Centered Design, and User Engagement and Gamification Dynamics through a five-phase structure supported by activities, artifacts, and stakeholder validation procedures. The study contributes a traceable Design Science-based development process that connects empirical findings with design decisions and provides a methodological foundation for the complete methodological specification and empirical evaluation of MIDA5. Accordingly, this study focuses on artifact construction rather than on demonstrating the effectiveness of the resulting methodology.

Information doi: 10.3390/info17070696

Authors: Chong Ho Yu Han Nee Chong

This paper examines how the emergence of over-parameterized artificial intelligence models and the phenomenon of double descent challenge the classical assumption that simpler models generalize better. Traditional predictive analytics relied on parsimonious models grounded in the bias-variance trade-off, where increasing complexity was expected to produce overfitting. However, recent advances in deep learning demonstrate that highly over-parameterized models can achieve superior generalization after surpassing the interpolation threshold. This paradigm shift has enabled systems such as AlphaFold, Aurora, Delphi-2M, and recommenders to model complex, high-dimensional relationships through contextual attention rather than global feature selection. The paper argues that higher education analytics remains largely reductionist, relying on limited variables such as GPA, demographics, and course completion rates to identify “at-risk” students. While interpretable, these approaches often fail to capture the dynamic and multidimensional nature of student success. In response, this study proposes a transition toward over-parameterized personalization, where students’ academic and behavioral histories are modeled as longitudinal high-dimensional sequences. Drawing parallels to commercial recommendation systems such as Amazon, Netflix, and YouTube, the paper explores how higher education can move from generalized early-warning systems toward adaptive “n-of-1” interventions. Importantly, the paper is conceptual rather than empirical: it develops a research agenda and a set of testable propositions, and it identifies the evaluation designs—temporally valid prediction protocols and causal intervention studies—by which the promise of over-parameterized personalization in higher education should be assessed before any claim of superiority can be made.

Information doi: 10.3390/info17070695

Authors: Mayte Toscano Domínguez Giacomo Martirano Alfonso Quarati Monica De Martino

The growing deployment of European Data Spaces requires more than advanced technical infrastructures; it also demands specialised human competencies to ensure interoperability, governance, and trust. This paper redefines organisational readiness as a key indicator of the transition from experimental deployments to large-scale, governed data exchange. Although technical solutions provide the necessary foundation, training is essential to ensure that the digital transition is supported by professionals capable of operating these ecosystems effectively. To address the research question, “Which competencies are required to enable data reuse within Data Spaces?”, with a specific focus on spatial data and Urban Digital Twins, this study proposes a hybrid methodology. This approach combines a top-down analysis of reference architectures (e.g., EIF and DSSC Blueprint) with a bottom-up diagnosis based on an expert-based qualitative survey. Preliminary results suggest that, although 58% of organisations identify technical integration between BIM and GIS as the main obstacle, 47% identify the shortage of specialised skills as the primary bottleneck preventing operational maturity. The analysis reveals a Maturity Plateau that hinders the transition from technical connectivity (Data Space Maturity Model Level 2) to full operational maturity (Level 3). To overcome this barrier, the study presents a convergence matrix based on the DIS4SME initiative, mapping knowledge gaps to a competency roadmap. This framework identifies urgent training needs in the semantic and legal domains, and provides an operational guide for supporting organisational maturity across Data Spaces and strengthening professional expertise.

Information doi: 10.3390/info17070692

Authors: Noviati Nurani Nilo Legowo

The transition from ISO/IEC 27001:2013 to ISO/IEC 27001:2022 is often approached as a certification and documentation update. However, for banking institutions with critical digital operations, the transition may create a distinct form of information security exposure when existing controls, evidence, ownership, and risk treatment decisions are not fully aligned with the revised control structure. This study examines transitional risk exposure in Bank XYZ, a banking institution with an established ISO/IEC 27001:2013-based Information Security Management System (ISMS). Using a qualitative single-case study design, the research analyzes ISMS documents, 96 information security risk records, Statement of Applicability records, control implementation evidence, ISO/IEC 27001:2013-to-ISO/IEC 27001:2022 control mapping materials, and stakeholder validation within the scope of core banking development and operations in data center and disaster recovery center environments. Risk exposure was assessed by comparing inherent and residual risks using a likelihood-impact matrix, while transition readiness was evaluated through control applicability, evidence adequacy, ownership clarity, supplier dependency, human factor readiness, and transition governance. The findings show that Bank XYZ implemented or provided evidence for 109 of 114 ISO/IEC 27001:2013 controls and reduced 96 inherent risks, consisting of 38 Moderate and 58 Moderate-to-High risks, into 13 Low and 83 Low-to-Moderate residual risks. Nevertheless, the transition review revealed that a mature residual risk position under ISO/IEC 27001:2013 does not automatically indicate readiness for ISO/IEC 27001:2022. Of 11 newly introduced controls, 2 were implemented, 4 were partially implemented, 4 were not implemented, and 1 was not applicable. The most significant transition gaps were found in threat intelligence, information deletion, data masking, and data leakage prevention. The study contributes by distinguishing transitional risk exposure from residual risk and by offering a control-level basis for prioritizing ISO/IEC 27001:2022 transition activities in banking institutions.

Information doi: 10.3390/info17070694

Authors: Abdalilah Alhalangy

Institutional artificial intelligence (AI) decision-support systems progressively evaluate cases, determine eligibility, and allocate resources; yet, predicted efficacy alone does not guarantee equity, contestability, or responsible utilization. Current research frequently considers fairness measures, explainability, human oversight, and organizational governance as rather distinct issues. This paper presents a traceable bias-auditing framework that amalgamates prediction, explanation, selective human review, and structured recording into a cohesive operational decision pathway. Through design science research, the artifact was exhibited in a controlled proof-of-concept utilizing 8000 synthetic institutional situations and historically biased data labels. The foundational classifier was a logistic regression model. Selective escalation is initiated by the proximity of boundaries, tension in explanation patterns, and the rules governing review priorities. Three situations were evaluated: baseline prediction, prediction with explanation alone, and comprehensive architecture with review and audit recording. Explanations enhanced reviewability but did not significantly alter fairness outcomes. The proposed architecture improved F1 from 0.781 to 0.795, reduced the demographic parity gap from 0.070 to 0.010, decreased the equal opportunity gap from 0.116 to 0.036, and improved audit completeness from 0.33 to 1.00, while escalating only 4.8% of cases for human review. The results indicate that explanations attain institutional significance solely when linked to procedural regulations and enduring records. The evaluation was simulation-based; thus, the results should be interpreted as proof-of-concept evidence rather than direct field validation.

Information doi: 10.3390/info17070693

Authors: May Issa Aldossary Hina Gull

Skin cancer is considered a deadly disease globally, and the timely identification of the disease may save human life. This research presents a CNN–Transformer-based fusion framework for automated multi-class skin lesion classification. This approach combines ResNet50 and Vision Transformer (ViT) to categorize skin lesions using the HAM10000 dataset. To assess their efficacy, a comparison with CNN and ViT models is also carried out. Seven classes of skin cancer are used for training the models, and class weighting is used to correct dataset asymmetry. According to the experimental dataset, the suggested hybrid framework shows improved performance over CNN and ViT, considering the accuracy (0.97) and macro-averaged F1-score (0.95). Furthermore, the efficiency of the suggested model is demonstrated by the fact that it delivers performance that is competitive with several existing approaches. Overall results indicate that hybrid CNN–Transformer architectures present a viable path for automated skin lesion categorization. Grad-CAM++ is integrated to enhance model understanding and promote medical confidence by enabling physicians to view visualizations that show the areas impacting the model’s conclusions. But there are still issues, including poor generalization, computational complexity, and a lack of external validation. Future research will concentrate on enhancing interpretability for practical implementation, integrating clinical information, and evaluating several datasets.

Information doi: 10.3390/info17070691

Authors: Weiqiong Wang Qidong Xu Tianyu Zhao Fang Fang

Accurate dynamic cost prediction is essential for budget optimization and risk mitigation in State Grid projects. However, traditional models and even recent deep learning approaches fall short, as they treat cost drivers independently, adopt simplistic concatenation that destroys sourcewise structure, or fail to handle irregularly sampled and partially missing multi-source data. This paper proposes a novel data-driven framework that integrates multi-source business data through a hierarchical tensor fusion mechanism and a hybrid spatiotemporal architecture. The problem is formalized as multivariate time-series prediction with irregular sampling and missing modalities. The framework comprises three synergistic innovations: a differentiable low-rank CANDECOMP/PARAFAC (CP) decomposition layer with adaptive attention weights that preserves cross-source structure while enabling compact dimensionality reduction; a spatiotemporal attention-based bidirectional gated recurrent unit (Bi-GRU) that captures long-range temporal dependencies; and a graph convolutional network (GCN) that explicitly learns interrelations among cost drivers, a capability absent in most existing forecasting methods. The entire system is trained end to end with a customized loss combining mean squared error, quantile loss, and temporal consistency regularization. Extensive experiments on three State Grid substation projects demonstrate that the proposed method outperforms state-of-the-art baselines by 12.7–18.4% in MAPE and maintains robust performance with up to 40% of data missing. These results confirm that explicitly modeling both temporal evolution and driver interdependencies within a unified fusion framework is the key to reliable cost forecasting in large-scale infrastructure projects.

Information doi: 10.3390/info17070690

Authors: Nursultan Nyssanov Leila Rzayeva Alisher Batkuldin Zhaksylyk Kozhakhmet

Kazakh automatic speech recognition (ASR) presents a persistent challenge for large-scale multilingual models. This paper presents a systematic evaluation of 27 ASR pipeline configurations (three ASR models × three VAD methods × three post-processing strategies) on the Kazakh Speech Dataset (KSD), examining the contribution of model fine-tuning, voice activity detection (VAD) preprocessing, and large language model (LLM) post-correction and benchmarking the resulting pipelines against two non-Whisper foundation models. Language-specific fine-tuning reduces Word Error Rate (WER) from 43.20% (generic Whisper-large-v3) to 11.88% (Kazakh fine-tuned Whisper-turbo), a 31.32-percentage-point absolute reduction (72.5% relative; p < 0.001, bootstrap test); the effect persists after controlling for model size (generic Whisper-large-v3-turbo, 18.92%, vs. the same architecture after fine-tuning, 11.88%; p < 0.001). VAD preprocessing consistently degrades performance. Zero-shot post-correction with general-purpose LLMs yields no benefit and adds substantial latency: Gemma-2-9B and Qwen2.5-7B raise WER by 5.5 and 7.2 percentage points at real-time factors of 0.52 and 0.30, and a larger 32B model still degrades accuracy (+10.8 points), indicating that scale is not the limiting factor. Among all systems evaluated, a larger multilingual foundation model, SeamlessM4T-v2 (9.72% WER), outperforms the fine-tuned Whisper, showing that for Kazakh model coverage matters more than pipeline engineering. Character-level error analysis identifies systematic confusion between Kazakh-specific and Russian Cyrillic characters as a dominant error source. These findings establish that, for Kazakh under the evaluated conditions, model choice dominates pipeline add-ons: fine-tuning is essential, VAD and zero-shot LLM correction consistently hurt, and a strong multilingual model sets the best result; we further discuss the extent to which these conclusions extend to typologically similar Kipchak-Turkic languages.

Information doi: 10.3390/info17070689

Authors: Sami Ul Haq Sheila Castilho Yvette Graham

Multimodal machine translation (MMT) aims to integrate visual context with textual data to improve the translation of ambiguous source text, such as the inclusion of an image as additional context. However, the evaluation of systems largely still relies on automatic metrics designed to evaluate text alone, and do not account for additional modalities during evaluation. The lack of dedicated MMT evaluation methods often results in inconsistent findings and creates uncertainty regarding the actual contribution of visual context in translation. In this work, we examine the performance of state-of-the-art trained and untrained evaluation metrics, particularly when comparing multimodal and text-only systems. Our evaluation focuses on the degree to which existing metrics are sensitive enough to distinguish between multimodal and text-only machine translation systems. We further investigate the potential for automatically generated image descriptions to serve as effective contextual signals for improving metric sensitivity to multimodal tasks. Our results show that incorporating such visual information into supervised metrics yields better alignment with human judgment. While all metrics successfully distinguished image-aware from image-agnostic systems on general test sets, both n-gram-based and embedding-based metrics struggled with respect to contrastive evaluation designed to capture context-dependent errors. Furthermore, we discuss how the presence of visual context may influence human evaluator judgment, observing that given the opportunity, human ratings are often substantially revised, further emphasizing the critical role of context in the evaluation of MMT.

Information doi: 10.3390/info17070688

Authors: Qirui Chen Kamisah Osman

Generative artificial intelligence (AI) is increasingly entering teacher education, yet evidence remains limited on its responsible integration into discipline-specific pedagogical preparation. This study examined whether an AI-supported Collaborative TPACK Competency Development module based on problem-based learning (CTD-PBL) was associated with greater pre–post gains in pre-service physics teachers’ self-reported technological pedagogical content knowledge (TPACK) and perceived collaborative problem-solving (CPS) processes. Informed by ADDIE, the 8-week module used DeepSeek as a bounded scaffold for collaborative lesson design, feedback, verification, and reflective revision while preserving teacher judgment. An intact-class quasi-experimental pre-test/post-test design involved 130 third-year pre-service physics teachers at a public university in western China. Two existing classes were randomly allocated at the class level to CTD-PBL or conventional instruction. Compared with the conventional group, the CTD-PBL group reported higher post-test TPACK scores (M = 4.04 vs. M = 3.40, p < 0.001, d = 1.02) and higher perceived CPS process scores (M = 3.62 vs. M = 3.05, p < 0.001, d = 0.88), with stronger pre–post gains in both outcomes. The findings provide a bounded curriculum design case showing how generative AI can be embedded in physics teacher education through problem-based tasks, collaborative scaffolding, and human verification procedures.

Information doi: 10.3390/info17070687

Authors: Ao Han Tongyu Zhao Yanjun Pei Haining Chen Jun Zhou Hailei Yuan Pan Hu

Robust monitoring of weld pool dynamics is critical for automated arc welding; however, single-modality sensors are frequently constrained by severe optical interference and high-frequency environmental noise. To address these limitations, we propose M2WPR-Net, a novel multimodal framework that synergizes visual and acoustic signals for simultaneous weld width regression and physical quality classification. The architecture employs a dual-stream ResNet50 backbone to process heterogeneous sensory data. Specifically, the visual stream utilizes a Convolutional Block Attention Module (CBAM) to suppress intense arc glare and localize the weld pool. Concurrently, the acoustic stream transforms 1D audio sequences into 2D Gramian Angular Summation Field (GASF) textures, which are subsequently refined by Squeeze-and-Excitation (SE) networks to isolate target frequency channels. A central contribution of this study is a bidirectional cross-modal attention mechanism based on Query–Key–Value (Q-K-V) matrix operations. Overcoming the shortcomings of static feature concatenation, this module dynamically aligns the modalities, enabling acoustic cues to guide visual feature extraction and vice versa, thereby mitigating information bottlenecks. Optimized via a joint multi-task loss function, the proposed M2WPR-Net significantly outperforms existing single-modal and conventional fusion baselines. Experimental results demonstrate that the network achieves a Mean Absolute Error (MAE) of 0.18 mm for width prediction and a 93.5% accuracy in penetration state classification, confirming its resilience and practical applicability in complex industrial welding environments.

Information doi: 10.3390/info17070686

Authors: Zhenyang Hu Zhandong Liu Ruixia Song Ke Li Shuping Chen Zhihua Wang Yong Li Xiangwei Qi

To address the reliance on manual calibration and the performance degradation caused by loosely coupled modules in lane-wise vehicle counting under complex traffic scenarios, this paper presents a lane-wise vehicle counting system based on adaptive lane partitioning and multi-module integration. The system first applies YOLOPv2 for lane-line detection, providing the basis for lane-region partitioning. Subsequently, Hue-Saturation-Value (HSV) color segmentation, morphological processing, and contour filtering are employed to enhance the robustness of lane feature extraction. Leveraging perspective geometry, lane regions are constructed to achieve adaptive lane partitioning. For vehicle analysis, YOLOv11 is utilized for vehicle detection, and ByteTrack is adopted for multi-object tracking. These modules are combined with lane assignment to form an integrated pipeline that preserves trajectory continuity and mitigates identity loss under occlusion and motion blur. Furthermore, a PyQt5-based interactive visualization interface is developed to support video processing, real-time display, lane-region visualization, and statistical analysis of per-lane traffic flow and lane-change behaviors. Experimental results demonstrate the effectiveness and practicality of the proposed system in complex multi-lane traffic scenarios.

Information doi: 10.3390/info17070685

Authors: Gianfranco Bilardi Michele Schimd

The edit distance is a metric of dissimilarity between strings, widely applied in computational biology, speech recognition, and machine learning. Let ek(n) denote the average edit distance between random, independent strings of n characters from an alphabet of size k. This paper is concerned with the efficient computation of αk(n)=ek(n)/n as well as of αk=limn→∞αk(n). It is shown that αk(n)−Θ(logn/n)≤αk≤αk(n) and that αk is computable. An algorithm for the exact computation of αk(n) is presented with running time T=O(n2kmin(3n,kn)) thus, of limited practical use. An analysis of Monte Carlo estimates, based on McDiarmid’s inequality, shows how αk(n) can be evaluated with good accuracy and high confidence level, for rather large values of n. In particular, 99.9% confidence intervals of width approximately 10−2 are obtained for αk. An efficiently computable lower bound βk* to αk is derived, with limk→∞βk*=1. For k greater than a few dozens, significant bounds on αk can be obtained faster via βk* than by statistical methods. The above techniques yield improvements on most previous numerical estimates as well as results for alphabet sizes and string lengths not reported before.

Information doi: 10.3390/info17070684

Authors: Bingxu Cao Lu Zhao Hao Zhao Hao Xi Laixiang Xu Shengyuan Yang

Rapid and accurate fall detection for college students plays a crucial role in enhancing the level of intelligent safety management and improving emergency response mechanisms in universities. Existing models often struggle to achieve a satisfactory balance between efficiency and robustness. To address this, we propose an improved YOLOv12 as the baseline and introduce weighted convolution (WConv) to reduce computational complexity. Additionally, we integrate channel reduction attention (CRA) to strengthen the model’s anti-interference capability. Experimental results demonstrate that the proposed method achieves a mAP@50 of 99.5% and a mAP@[0.5:0.95] of 92.38% on our dataset, with a single-frame inference time of 443 ms. The proposed model strikes a well-balanced and effective trade-off between accuracy and speed, making it particularly suitable for deployment on embedded terminals and offering a reliable monitoring solution for physical education instruction.

Information doi: 10.3390/info17070683

Authors: Yutao Wang Zhengzheng Zhu Yongqiang Bai Zhibo Xie Renwei Tu

In the field of road damage detection, the accuracy of existing methods still requires further improvement, particularly for elongated cracks, which are crucial for ensuring driving safety and effective road maintenance. To address this limitation, a novel road damage detection algorithm is proposed based on direction awareness and feature equalization. Specifically, a Direction-aware Strip Convolution (DSC) module is constructed to effectively capture the geometric characteristics of elongated cracks and maintain computational efficiency, by integrating asymmetric strip convolution and depthwise separable convolution respectively. In addition, a Multi-level Feature Equalization (MFE) module is designed to address the complex morphology and significant scale variations of road damage during multi-level feature fusion. Specifically, a set of learnable spatial weighting parameters is introduced in this module, whose weighting coefficients are optimized across different network layers and adaptively generated, thereby modulating the contributions of multi-level features and promoting a more balanced multi-level feature representation. Experimental results on the RDD2022-based experimental dataset demonstrate that the proposed method improves mAP@50 by 5.8 percentage points and recall by 5.9 percentage points, while achieving a processing speed of 122 FPS. Notably, the proposed method improves the detection performance of elongated cracks and achieves relatively balanced performance gains across different road damage categories, compared with the baseline model.

Information doi: 10.3390/info17070682

Authors: Mohammed M. Al-Gawda Majdi Abdellatief Ibrahim Al-Baltah

Despite widespread adoption of cybersecurity awareness training (CSAT), a persistent knowledge–behaviour gap continues to undermine organisational security posture, particularly in resource-constrained and developing-country contexts. This 12-week quasi-experimental field study evaluated an AI-adaptive CSAT platform against traditional instructor-led training (ILT) across three Yemeni organisations (total N = 187; AI-Adaptive: n = 94; Control: n = 93). The system used a 4-parameter Bayesian Knowledge Tracing (BKT) engine—with interpretable guess and slip signals—as an auditable pedagogical decision layer that triggered Protection Motivation Theory (PMT) and Theory of Planned Behavior (TPB)-aligned interventions. Extreme baseline imbalances (Cohen’s d > 2.0), at which standard ANCOVA residual adjustment alone is known to be biased and which necessitated advanced causal-inference triangulation, were addressed via a four-method protocol (ANCOVA, Propensity Score Matching, Difference-in-Differences, mixed-effects). All four methods converged on consensus effect sizes of d = 0.66–0.89. IT-verified Tier 2–3 incidents declined by 48.9% (incidence-rate ratio [IRR] = 0.51, 95% CI [0.38, 0.68]); blinded phishing click-rates fell from 8.8% to 2.1% (χ2(1) = 8.74, p = 0.003). Bootstrapped mediation analysis (PROCESS Model 4; 5000 draws) indicated that coping self-efficacy and perceived behavioural control—but not threat appraisal—were jointly associated with 66.4% of the total compliance effect. Rosenbaum bounds Γ = 2.1; E-values ≥ 3.4. The findings are consistent with the hypothesis that AI-adaptive cybersecurity training produces robust, theoretically explicable benefits and that the coping-appraisal pathway, not threat salience, is the active psychological mechanism. The four-method triangulation framework offers a replicable standard for field evaluations with non-random assignment. the consensus envelope d = 0.66–0.89 is the observed range of point estimates across the four estimators; per-method 95% CIs are reported below indirect effect via coping self-efficacy = 0.843 [0.52, 1.19], via PBC = 0.524 [0.28, 0.81], via threat appraisal = 0.059 [−0.07, 0.21] (ns); direct effect c’ = 0.63 (p = 0.026); total effect c = 2.06 [1.58, 2.54].

Information doi: 10.3390/info17070681

Authors: Zhanibek Balabayev Svitlana Biloshchytska Beibit Abdikenov Tomiris Zhaksylyk Birzhan Ayanbayev Dimash Rakishev

Many historical handwritten records in low-resource languages remain difficult to access through modern digital systems. This limits efforts to preserve and study cultural heritage at scale. Chagatai manuscripts exemplify these challenges within the Eastern Turki tradition. For centuries, it served as a major written language across Central Asia and supported a rich literary tradition. Large collections of Chagatai manuscripts still survive today, yet only a small amount of this material exists in digital form. As the technical literature specifically focused on Chagatai-HTR remains in its nascent stage, this review synthesizes indirect evidence from taxonomically related Perso-Arabic scripts to establish a foundational research framework. This article presents a systematic literature review following the PRISMA guidelines to examine artificial intelligence methods for handwritten text recognition (HTR) and text restoration in low-resource languages. Analyzing 50 studies published between 2020 and 2026, the review categorizes research trends into handwritten text recognition (HTR), optical character recognition (OCR), script classification, dataset development, and multimodal vision–language systems. The findings reveal a significant architectural shift from traditional segmentation-based CNN and RNN models toward transformer architectures and multimodal approaches. However, for Chagatai specifically, the primary obstacle is not the lack of advanced models but a critical scarcity of basic research infrastructure, including expert-verified transcriptions, annotation standards, and open benchmark datasets. Consequently, this article proposes a concrete development roadmap focusing on systematic digitization, expert annotation, transfer learning, and the creation of baseline models to enable reproducible evaluations.

Information doi: 10.3390/info17070680

Authors: Farai Fredric Mlambo Mpho Musuphi Kopano Letsela

Customer churn remains a significant concern in the telecommunications sector, leading to reduced profits and increased customer acquisition costs. The competitive nature of the industry allows customers the freedom to switch providers easily, necessitating effective models to predict and mitigate churn. This paper aimed to develop and compare optimal hybrid models for accurately predicting customer churn within a specified timeframe and to assess how various factors influence the time until churn. To achieve this, a dataset from an Iranian telecommunications company was utilised. The methodology involved a three-stage hybrid approach: initially, customers were segmented using K-means (KM) and Agglomerative clustering (AC) techniques. Subsequently, binary classification was performed using Logistic Regression (LR), Artificial Neural Networks (ANN), and Decision Trees (DT). Finally, the Cox proportional hazard model (CoxPH) was employed to estimate hazard rates and analyse the impact of covariates on churn time. Model performance was evaluated using metrics such as Accuracy, Precision, Recall, F1-Score, Specificity, and Area Under the Receiver Operating Characteristic Curve (AUC-ROC). The experimental results demonstrated that hybrid models generally outperformed individual Cox models across most predictive performance metrics. Specifically, the K-means + Logistic Regression + Cox (KM + LR + Cox) model was identified as the best performer, achieving an Accuracy of 88.24%, Precision of 93.18%, Specificity of 63.64%, and an F1-score of 93.01%. KM with two clusters (representing churners and non-churners) was optimal for customer segmentation. Covariate analysis revealed that factors such as ‘Cluster 1’ decreased survival length, suggesting that customers in ‘Cluster 2’ are more prone to churn. This paper successfully developed an optimal three-stage hybrid model, KM + LR + Cox, which effectively predicts customer churn and identifies key factors influencing churn duration, offering valuable insights for targeted retention strategies.

Information doi: 10.3390/info17070678

Authors: Minling Zeng Wenfeng Liu Yanling Zheng

Although monetization has been widely discussed in digital games, less attention has been paid to how it is discussed in narrative-centered Interactive Full-Motion Video (FMV) Games, where user experience depends heavily on continuity, completeness, and emotional engagement. Addressing this gap, this study analyzes 8676 Steam user reviews of Road to Empress I using NLP-based preprocessing, semantic analysis, sentiment analysis, and LDA topic modeling. The results indicate that the review corpus was organized around several interconnected dimensions, including narrative progression, character engagement, audiovisual performance, interactivity, and payment structure, with narrative-related concerns occupying a central position. Compared with the full corpus, the identified monetization-related review subset showed a less positive sentiment distribution, suggesting that monetization-related discussion within this subset was associated with more negative evaluations. Topic modeling further indicated three prominent concerns: interrupted narrative continuity, perceived incompleteness, and dissatisfaction with value for money. Overall, the findings suggest that, within the identified subset, monetization was discussed not merely as a pricing issue, but as part of the broader experiential structure through which continuity, completeness, and value were judged. The study contributes to research on Interactive FMV Games by foregrounding monetization as an experiential dimension and offers practical implications for designing less disruptive monetization strategies in narrative-centered interactive products.

Information doi: 10.3390/info17070679

Authors: Bilgin Metin Nazlı Elif Yey Martin Wynn

AI-driven digitalisation transforms how organisations process personal data and introduces risks that traditional Data Protection Impact Assessment (DPIA) frameworks cannot adequately address. Automated decision-making and large-scale processing in AI, IoT, big data analytics, and blockchain environments create privacy concerns beyond the scope of existing DPIA methodologies. The EU AI Act extends this scope through the Fundamental Rights Impact Assessment (FRIA) under Article 27, which links data protection obligations to broader fundamental rights governance. This study addresses these gaps through a two-phase research design. Phase 1 conducts a systematic literature review of 25 studies and applies framework analysis to identify DPIA implementation challenges across four categories: legal and regulatory, risk assessment, scope, and complexity. AI-specific challenges appear across all four categories. Phase 2 develops a governance framework built on the Plan-Do-Check-Act (PDCA) cycle and organised through a four-level hierarchy of Lifecycle Phase, Risk Management Domain, Control Objective, and Operational Activity. The framework translates relevant requirements of ISO 31000:2018, ISO/IEC 27701:2025, and ISO/IEC 29134:2023 into traceable activities and encompasses algorithmic fairness and socio-ethical impacts. The actionable DPIA framework supports compliance with the GDPR, the EU AI Act and the three ISO standards and will be of interest to company practitioners and other researchers investigating the theoretical and practice-based aspects of digitalisation and data privacy.

Information doi: 10.3390/info17070677

Authors: Shanshan Peng Fang Zhu

As generative artificial intelligence (GenAI) increasingly permeates educational contexts, understanding the factors driving learners’ Behavioral Intention (BI) toward GenAI-powered tools has become critical. This study integrates the Technology Acceptance Model (TAM), the Task-Technology Fit (TTF) framework, and privacy and ethical risk considerations to explore the determinants of Chinese higher education students’ Behavioral Intention to adopt these tools. Data were collected from 716 students via a structured self-reported questionnaire. A multi-stage analytical approach was employed by integrating structural equation modeling (SEM) with artificial neural networks (ANN) and support vector regression (SVR). SEM was first utilized to validate the theoretical hypotheses and the measurement model. Subsequently, ANN and SVR models were constructed to explore non-linear relationships and rank the importance of core predictors for Behavioral Intention, including Perceived Ease of Use (PEU), Privacy and Ethical Concerns (PEC), Perceived Technical Features (PTF), and TTF. The modeling performance of the two algorithms was then rigorously compared. The SEM results indicate that PTF exerts an indirect impact on Behavioral Intention via the sequential mediation of Task-Technology Fit and Perceived Usefulness (PU), while PEU positively influences both Perceived Usefulness and Behavioral Intention. Notably, PEC did not exhibit a significant negative effect on users’ Attitude (ATT) or Behavioral Intention. These findings were further elucidated by the machine learning analyses, where PTF and PEU emerged as the dominant predictors, whereas the non-linear contribution of PEC was marginal. Furthermore, SVR outperformed ANN in terms of predictive accuracy and model stability. This study demonstrates the efficacy of combining theoretical modeling with machine learning techniques to elucidate the adoption mechanisms of GenAI in higher education. In addition, preliminary teaching observations in undergraduate mathematics and logistics management courses link quantitative results with actual learning scenarios. We acknowledge that future research should validate these patterns using observed behavioral data.

Information doi: 10.3390/info17070676

Authors: Umar Adeel Ammar Rashid Shafiz Affendi Bin Mohd Yusof Usman Javed Butt

Vehicle theft and vandalism remain significant urban security challenges commonly addressed through reactive, post-incident forensic measures. This paper proposes a proactive, real-time computer vision system designed to detect potentially suspicious behavior around parked vehicles, with a specific focus on unauthorized proximity and loitering. The proposed architecture integrates state-of-the-art object detection using YOLOv11 (You Only Look Once version 11), multi-object tracking via a lightweight custom association tracker inspired by the ByteTrack/StrongSORT/OC-SORT paradigm, and monocular depth estimation based on the Intel DPT-Large framework.A key contribution is the identification and mitigation of the Perspective Challenge: the two-dimensional (2D) scale ambiguity that causes distant background pedestrians to appear falsely proximate to foreground vehicles in monocular camera feeds. To address this, three spatial analysis strategies are implemented and evaluated: (A) fixed Euclidean thresholding, (B) adaptive perspective thresholding, and (C) three-dimensional (3D) depth injection. Experimental results on real-world urban surveillance footage (27,000 annotated frames across two datasets) demonstrate that Strategy C achieves the highest precision (0.95) with an F1-score of 0.92, while Strategy B provides the best balance between accuracy (precision 0.88, recall 0.91, F1 0.89) and computational efficiency (32.7 frames per second, FPS). Compared to naive 2D thresholding (Strategy A), Strategy B reduces false alarms by approximately 80%, while Strategy C further improves precision to 0.95 through depth-plane verification. The system maintains real-time performance exceeding 30 FPS under Strategy B, making it a strong candidate for practical urban vehicle monitoring, subject to further large-scale validation across diverse environments.

Information doi: 10.3390/info17070675

Authors: Liwen Zhang Yiqi Liu Jingya Li Yuexi Dong

This paper proposes the AI–Human Collaborative Interpretation (AHCI) human–machine collaboration methodological framework, offering a new research pathway for the digital interpretation of complex cultural heritage. The framework integrates multi-source data processing, formalised feature modelling, interpretable analysis and human–machine feedback mechanisms, translating the complex information of cultural heritage into an analysable and interactive structure. Building upon this, the Tianlai.China project serves as a case study for the multimodal data integration, aesthetic feature modelling and interactive visualisation of the Illustrated Catalogue of Famous Porcelains Through the Ages. The effectiveness of the collaborative mechanism is validated through expert–public–AI comparative experiments and human–AI feedback loop experiments. The main contributions of this paper include: proposing a methodological framework for human–machine collaboration in cultural heritage value assessment; constructing an interpretable formalised model of cultural features and a corresponding visualisation system; analysing the impact of AI explanations on expert cognition through empirical research; and exploring the transferability of this framework across different types of cultural heritage. Against this backdrop, the human–machine collaboration system developed by AHCI—comprising ‘data semantics–feature modelling–visual output–cognitive feedback’—possesses cross-domain applicability.

Information doi: 10.3390/info17070674

Authors: Bogdan Arct Mateusz Bieniek Bartłomiej Kanabus Aleksander Kozłowski Piotr Wetmański Michał Kruk Sylwia Stachowiak Jarosław Kurek

Large online marketplaces increasingly rely on multi-stage ranking pipelines where a learned relevance model is complemented by business-aware constraints such as contractual pacing, exposure caps and commercial alignment objectives. This paper develops a second-stage, contract-aware re-ranking layer for real-estate recommendation that explicitly balances user–item relevance with plan fulfillment, lead value and operational guardrails. The proposed multi-objective re-ranker (PMOR) combines a calibrated base relevance score with multiplicative business adjustments and subtractive penalties for approaching contractual caps and for within-slate similarity. The method supports heterogeneous settlement models, including pay-per-action and fixed-fee contracts, via contract-specific weights. Because the scoring function is deterministic and structured, it admits exact component-wise contribution analysis and counterfactual ablations without relying on surrogate explainability methods. Offline evaluation on production logs from an anonymized marketplace covers 4219 recommendation requests and 184,147 candidate items, joined with daily business snapshots using an as-of strategy to prevent look-ahead bias. Under a profit proxy based on effective lead value, position discounting and billability, PMOR achieves an indexed expected-revenue proxy of 487.4 (baseline = 100), corresponding to a lift of 387.4% over a model-only baseline and 48.4% over a legacy production re-ranker (LPR). The gain is primarily associated with improved billable exposure, increasing the share of billable positions in TOP-3 to 78.22% compared with 37.85% for LPR. We discuss parameter sensitivity, operational considerations and limitations of offline proxy objectives for deployment.

Information doi: 10.3390/info17070673

Authors: Zhiyuan Li Songsong Li Weining Li Meide Liu Shidong Li Xiaoming Chen

Weld defect detection is challenged by significant variations in defect scale, complex defect morphology, and strong background interference. To address these issues, this study proposes WID-YOLO11, an improved detection model based on YOLO11n. First, a novel C3k2_LFA-WTConv module is developed to decompose feature maps into frequency subbands via learnable wavelet filters and adaptively reweight them through a Learnable Frequency-Attention (LFA) mechanism, enabling the network to learn task-adapted subband representations for different defect types, where each class exhibits distinct wavelet subband energy distributions. Second, an Improved Multi-Scale Dilated Local Attention (IMSDA) module is constructed by extending the dilation-rate set of the original MSDA from three to four values (1, 2, 3, and 4), expanding the maximum equivalent receptive field from 7 × 7 to 9 × 9 to improve multi-scale sensitivity to medium-scale defects such as porosity and spatter at low computational overhead. Finally, a DySample dynamic upsampling module is integrated into the neck network to adaptively learn sampling locations and preserve defect boundary details more effectively during multi-scale feature fusion. Evaluations on a publicly available weld defect dataset of 986 images show that WID-YOLO11 outperforms the YOLO11n baseline by 4.5, 10.4, and 3.4 percentage points in mAP@0.5, Recall, and mAP@0.5:0.95, respectively, confirming the effectiveness of the three proposed enhancements.

Information doi: 10.3390/info17070672

Authors: Wipada Sriwichien Warawut Narkbunnum Kittipol Wisaeng

Artificial intelligence (AI) adoption within logistics organizations remains uneven despite increasing digital transformation initiatives in emerging economies. This study investigates respondent-perceived organizational AI readiness profiles in Thailand’s logistics industry using an integrated analytical framework combining TOE–UTAUT predictors, clustering analysis, supervised machine learning, and explainable artificial intelligence techniques. Data were collected from 520 logistics and supply chain professionals in Thailand using a structured questionnaire. K-means clustering was applied to identify internally derived respondent-perceived AI readiness profiles, while Random Forest, Support Vector Machine (SVM), XGBoost, and LightGBM models were developed to classify readiness-profile membership. A weighted voting ensemble model was additionally employed to assess classification robustness and profile-differentiation stability across multiple learning algorithms. The findings identified three internally derived respondent-perceived AI readiness profiles representing relatively low, moderate, and advanced readiness patterns within the TOE–UTAUT feature space. Among the evaluated models, the SVM classifier achieved the strongest classification performance, obtaining the highest accuracy and AUC values. SHAP analysis indicated that Actual Use, Technological Factors, Facilitating Conditions, and Behavioral Intention exhibited the largest feature-attribution contributions within the readiness-profile classification framework. The study contributes to AI adoption research by integrating clustering-based segmentation, machine-learning classification, and explainable artificial intelligence into a unified readiness-profiling framework. The findings provide practical insights for managers and policymakers seeking to understand respondent-perceived organizational AI readiness patterns and support digital transformation initiatives within logistics professional contexts.

Information doi: 10.3390/info17070671

Authors: Jia Chen Yiheng Ma Shijuan Gao

Legal mediation is an important mechanism for resolving social conflicts and handling disputes. It involves complex interpersonal interactions and unstructured decision-making processes, and therefore holds significant research value as a domain. Leveraging the outstanding logical reasoning capabilities of large language models, multi-agent systems for simulating complex social interactions have become a cutting-edge research direction in artificial intelligence, providing a new supporting vehicle and research pathway for the intelligent study and practical application of legal mediation. However, directly applying general-purpose multi-agent techniques or general-purpose, opaque LLMs to long-horizon, multi-party, and high-conflict professional mediation tasks exposes several deep-seated structural cognitive deficiencies, including a lack of process awareness, insufficient domain-specific intervention capabilities, and limited theory-of-mind reasoning. To address these challenges, this study proposes CogMed, a cognitively enhanced multi-agent framework for legal mediation simulation, which aims to compensate for the limitations of general models in professional strategic interactions through an explicit cognitive architecture. Rather than introducing entirely new individual reasoning modules, the proposed framework focuses on cognitively coordinated integration of process control, strategic intervention, and belief modeling mechanisms under legal mediation settings. CogMed models the mediation process as a Finite State Machine (FSM) to capture macro-level decision logic and introduces a Strategic Toolkit (STK) that serves as a set of action primitives for micro-level interventions. Meanwhile, a Dynamic Belief Tracking (DBT) mechanism is incorporated into party agents to simulate psychological anticipation and strategic reasoning during negotiation. Experimental results demonstrate that CogMed effectively improves both mediation success rates and the quality of negotiated outcomes. Furthermore, the findings suggest a preliminary framework-level compensation pattern under the current experimental setting, where cognitively structured coordination mechanisms may partially enhance the mediation capability of medium-scale models. These preliminary experimental observations suggest that cognitively structured coordination mechanisms may partially compensate for certain limitations associated with model scale under the current controlled mediation setting, thereby offering a potential research direction for cognitively structured legal mediation simulation systems under controlled experimental settings.