Big Data Cogn. Comput., Volume 9, Issue 11 (November 2025) – 31 articles

Educational equity and access to quality learning opportunities represent fundamental pillars of sustainable societal development, directly aligned with the United Nations Sustainable Development Goal 4 (Quality Education). Student retention remains a critical challenge in higher education, with early disengagement strongly predicting eventual failure and limiting opportunities for social mobility. While machine learning models have demonstrated impressive predictive accuracy for identifying at-risk students, most systems prioritize performance metrics over practical deployment constraints, creating a gap between research demonstrations and real-world impact for social good. We present an accountable and interpretable decision support system that balances three competing objectives essential for responsible AI deployment: ultra-early prediction timing (day 14 of semester), manageable instructor workload (flagging 15% of students), and model transparency (multiple explanation mechanisms). Using the Open University Learning Analytics Dataset (OULAD) containing 22,437 students across seven modules, we develop predictive models from activity patterns, assessment performance, and demographics observable within two weeks. We compare threshold-based rules, logistic regression (interpretable linear modeling), and gradient boosting (ensemble modeling) using temporal validation where early course presentations train models tested on later cohorts. Results show gradient boosting achieves AUC (Area Under the ROC Curve, measuring discrimination ability) of 0.789 and average precision of 0.722, with logistic regression performing nearly identically (AUC 0.783, AP 0.713), revealing that linear modeling captures most predictive signal and makes interpretability essentially free. At our recommended threshold of 0.607, the predictive model flags 15% of students with 84% precision and 35% recall, creating actionable alert lists instructors can manage within normal teaching duties while maintaining accountability for false positives. Calibration analysis confirms that predicted probabilities match observed failure rates, ensuring trustworthy risk estimates. Feature importance modeling reveals that assessment completion and activity patterns dominate demographic factors, providing transparent evidence that behavioral engagement matters more than student background. We implement a complete decision support system generating instructor reports, explainable natural language justifications for each alert, and personalized intervention templates. Our contribution advances responsible AI for social good by demonstrating that interpretable predictive modeling can support equitable educational outcomes when designed with explicit attention to timing, workload, and transparency—core principles of accountable artificial intelligence. Full article

Romanian popular music has had a storied history across the last century and a half. Incorporating different influences at different times, today it boasts a wide range of both autochthonous and imported genres, such as traditional folk music, rock, rap, pop, and manele, to name a few. We aim to trace the linguistic differences between the lyrics of these genres using natural language processing and a computational linguistics approach by studying the prosodic, stylistic, syntactic, and sentiment-based features of each genre. For this purpose, we have crawled a dataset of ~14,000 Romanian songs from publicly available websites along with the user-provided genre labels, and characterized each song and each genre, respectively, with regard to these features, discussing similarities and differences. We improve on existing tools for Romanian language natural language processing by building a lexical analysis library well suited to song lyrics or poetry which encodes a set of 17 linguistic features. In addition, we build lexical analysis tools for profanity-based features and improve the SentiLex sentiment analysis library by manually rebalancing its lexemes to overcome the limitations introduced by it having been machine translated into Romanian. We estimate the accuracy gain using a benchmark Romanian sentiment analysis dataset and register a 25% increase in accuracy over the SentiLex baseline. The contribution is meant to describe the characteristics of the Romanian expression of autochthonous as well as international genres and provide technical support to researchers in natural language processing, musicology or the digital humanities in studying the lyrical content of Romanian music. We have released our data and code for research use. Full article

Magnetic Resonance Imaging (MRI) reconstruction is computationally heavy from under sampled data, and centralized data sharing within deep learning models was met with privacy concerns. We therefore propose FL-Swarm MRCM, a novel federated learning framework that integrates FedDyn dynamic regularization, a swarm-optimized generative adversarial network (SwarmGAN), and a structure-aware cross-entropy loss to enhance cross-site MRI reconstruction without sharing raw data. The framework avoids client drift, locally adapts hyper-parameters using Particle Swarm Optimization, and preserves anatomic fidelity. Evaluations on fastMRI, BraTS-2020, and OASIS datasets under non-IID partitions show that FL-Swarm MRCM improves reconstruction quality, achieving PSNR = 29.78 dB and SSIM = 0.984, outscoring FL-MR and FL-MRCM baselines. The federated framework for adversarial training proposed here enables reproducible, privacy-preserving, and strongly multi-institutional MRI reconstruction with better cross-site generalization for clinical use. Full article

Cloud resource provider deployment at random locations increases operational costs regardless of the application demand intervals. To provide adaptable load balancing under varying application traffic intervals, the auto-scaling concept has been introduced. This article introduces a Pervasive Auto-Scaling Method (PASM) for Computing Resource Allocation (CRA) to improve the application quality of service. In this auto-scaling method, deep reinforcement learning is employed to verify shared instances of up-scaling and down-scaling pervasively. The overflowing application demands are computed for their service failures and are used to train the learning network. In this process, the scaling is decided based on the maximum computing resource allocation to the demand ratio. Therefore, the learning network is also trained using scaling rates from the previous (completed) allocation intervals. This process is thus recurrent until maximum resource allocation with high sharing is achieved. The resource provider migrates to reduce the wait time based on the high-to-low demand ratio between successive computing intervals. This enhances the resource allocation rate without high wait times. The proposed method’s performance is validated using the metrics resource allocation rate, service delay, allocated wait time, allocation failures, and resource utilization. Full article

Background: Understanding visual context in images is essential for enhanced Point-of-Interest (POI) recommender systems. Traditional models often rely on global features, overlooking object-level information, which can limit contextual accuracy. Methods: This study introduces micro-level contextual tagging, a method for extracting metadata from individual objects in images, including object type, frequency, and color. This enriched information is used to train WORLDO, a Vision Transformer model designed for multi-task learning. The model performs scene classification, contextual tag prediction, and object presence detection. It is then integrated into a content-based recommender system that supports feature configurations. Results: The model was evaluated on its ability to classify scenes, predict tags, and detect objects within images. Ablation analysis confirmed the complementary role of tag, object, and scene features in representation learning, while benchmarking against CNN architectures showed the superior performance of the transformer-based model. Additionally, its integration with a POI recommender system demonstrated consistent performance across different feature settings. The recommender system produced relevant suggestions and maintained robustness even when specific components were disabled. Conclusions: Micro-level contextual tagging enhances the representation of scene context and supports more informative recommendations. WORLDO provides a practical framework for incorporating object-level semantics into POI applications through efficient visual modeling. Full article

Visual object tracking is a fundamental computer vision task recently extended to multimodal settings, where natural language descriptions complement visual information. Existing multimodal trackers typically rely on large-scale transformer architectures that jointly train visual and textual encoders, resulting in hundreds of millions of trainable parameters and substantial computational overhead. We propose a lightweight multimodal adapter that integrates textual descriptions into a state-of-the-art visual-only framework with minimal overhead. The pretrained visual and text encoders are frozen, and only a small projection network is trained to align text embeddings with visual features. The adapter is modular, can be toggled at inference, and has negligible impact on speed. Extensive experiments demonstrate that textual cues improve tracking robustness and enable efficient multimodal integration with over 100× fewer trainable parameters than heavy multimodal trackers, allowing training and deployment on resource-limited devices. Full article

Natural Language Processing is being used for Disease Outbreak Prediction using news data. However, the available research focuses on predicting outbreaks for only specific diseases using disease-specific data such as COVID-19, Zika, SARS, MERS, and Ebola, etc. To address the challenge of disease outbreak prediction without relying on prior knowledge or introducing bias, this research proposes a model that leverages a news dataset devoid of specific disease names. This approach ensures generalizability and domain independence in identifying potential outbreaks. To facilitate supervised learning, spaCy was employed to annotate the dataset, enabling the classification of articles as either related or unrelated to disease outbreaks. LSTM, Bi-LSTM, and Bi-LSTM with a Multi-Head Attention mechanism, and transformer have been used and compared for the purpose of classification. Experimental results exhibit good prediction accuracy with Bi-LSTM with Multi-Head Attention and transformer on the test dataset. The work serves as a pro-active and unbiased approach to predict any disease outbreak without being specific to any disease. Full article

Wildfires pose a growing threat to ecosystems, infrastructure, and public safety, particularly in the province of British Columbia (BC), Canada. In recent years, the frequency, severity, and scale of wildfires in BC have increased significantly, largely due to climate change, human activity, and changing land use patterns. This study presents a comprehensive, data-driven approach to wildfire prediction, leveraging advanced machine learning (ML) and deep learning (DL) techniques. A high-resolution dataset was constructed by integrating five years of wildfire incident records from the Canadian Wildland Fire Information System (CWFIS) with ERA5 reanalysis climate data. The final dataset comprises more than 3.6 million spatiotemporal records and 148 environmental, meteorological, and geospatial features. Six feature selection techniques were evaluated, and five predictive models—Random Forest, XGBoost, LightGBM, CatBoost, and an RNN + LSTM—were trained and compared. The CatBoost model achieved the highest predictive performance with an accuracy of 93.4%, F1-score of 92.1%, and ROC-AUC of 0.94, while Random Forest achieved an accuracy of 92.6%. The study identifies key environmental variables, including surface temperature, humidity, wind speed, and soil moisture, as the most influential predictors of wildfire occurrence. These findings highlight the potential of data-driven AI frameworks to support early warning systems and enhance operational wildfire management in British Columbia. Full article

Speech separation, as an important research direction in audio signal processing, has been widely studied by the academic community since its emergence in the mid-1990s. In recent years, with the rapid development of deep neural network technology, speech processing based on deep neural networks has shown outstanding performance in speech separation. While existing studies have surveyed the application of deep neural networks in speech separation from multiple dimensions including learning paradigms, model architectures, loss functions, and training strategies, current achievements still lack systematic comprehension of the field’s developmental trajectory. To address this, this paper focuses on single-channel supervised speech separation tasks, proposing a technological evolution path “U-Net–TasNet–Transformer–Mamba” as the main thread to systematically analyze the impact mechanisms of core architectural designs on separation performance across different stages. By reviewing the transition process from traditional methods to deep learning paradigms and delving into the improvements and integration of deep learning architectures at various stages, this paper summarizes milestone achievements, mainstream evaluation frameworks, and typical datasets in the field, while also providing prospects for future research directions. Through this detailed-focused review perspective, we aim to provide researchers in the speech separation field with a clearly articulated technical evolution map and practical reference. Full article

Understanding and accurately estimating cognitive workload is crucial for the development of adaptive, user-centered interactive systems across a variety of domains including augmented reality, automotive driving assistance, and intelligent tutoring systems. Cognitive workload assessment enables dynamic system adaptation to improve user experience and safety. In this work, we introduce a novel framework that leverages smooth pursuit eye movements as a non-invasive and temporally precise indicator of mental effort. A key innovation of our approach is the development of trajectory-independent algorithms that address a significant limitation of existing methods, which generally rely on a predefined or known stimulus trajectory. Our framework leverages two solutions to provide accurate cognitive load estimation, without requiring knowledge of the exact target path, based on Kalman filter and B-spline heuristic classifiers. This enables the application of our methods in more naturalistic and unconstrained environments where stimulus trajectories may be unknown. We evaluated these algorithms against classical supervised machine learning models on a publicly available benchmark dataset featuring diverse pursuit trajectories and varying cognitive workload conditions. The results demonstrate competitive performance along with robustness across different task complexities and trajectory types. Moreover, our framework supports real-time inference, making it viable for continuous cognitive workload monitoring. To further enhance deployment feasibility, we propose a federated learning architecture, allowing privacy-preserving adaptation of models across heterogeneous devices without the need to share raw gaze data. This scalable approach mitigates privacy concerns and facilitates collaborative model improvement in distributed real-world scenarios. Experimental findings confirm that metrics derived from smooth pursuit eye movements reliably reflect fluctuations in cognitive states induced by working memory load tasks, substantiating their use for real-time, continuous workload estimation. By integrating trajectory independence, robust classification techniques, and federated privacy-aware learning, our work advances the state of the art in adaptive human–computer interaction. This framework offers a scientifically grounded, privacy-conscious, and practically deployable solution for cognitive workload estimation that can be adapted to diverse application contexts. Full article

Person–job data are multi-source, heterogeneous, and strongly temporal, making knowledge modeling and analysis challenging. We present an automated approach for constructing a Human-Resources Temporal Knowledge Graph. We first formalize a schema in which temporal relations are represented as sets of time intervals. On top of this schema, a large language model (LLM) pipeline extracts entities, relations, and temporal expressions, augmented by self-verification and external knowledge injection to enforce schema compliance, resolve ambiguities, and automatically repair outputs. Context-aware prompting and confidence-based escalation further improve robustness. Evaluated on a corpus of 2000 Chinese resumes, our method outperforms strong baselines, and ablations confirm the necessity and synergy of each component; notably, temporal extraction attains an F1 of 0.9876. The proposed framework provides a reusable path and engineering foundation for downstream HR tasks—such as profiling, relational reasoning, and position matching—supporting more reliable, time-aware decision-making in complex organizations. Full article

Automated violence detection in video surveillance is critical for public safety; however, existing methods frequently suffer notable performance degradation across diverse real-world scenarios due to domain shift. Substantial distributional discrepancies between source training data and target environments severely hinder model generalization, limiting practical deployment. To overcome this, we propose CoMT-VD, a new contrastive Mean Teacher-based violence detection model, engineered for enhanced adaptability in unseen target domains. CoMT-VD innovatively integrates a Mean Teacher architecture to adequately leverage unlabeled target domain data, fostering stable, domain-invariant feature representations by enforcing consistency regularization between student and teacher networks, crucial for bridging the domain gap. Furthermore, to mitigate supervisory noise from pseudo-labels and refine the feature space, CoMT-VD incorporates a dual-strategy contrastive learning module. DCL systematically refines features through intra-sample consistency, minimizing latent space distances for compact representations, and inter-sample consistency, maximizing feature dissimilarity across distinct categories to sharpen decision boundaries. This dual regularization purifies the learned feature space, boosting discriminativeness while mitigating noisy pseudo-labels. Broad evaluations on five benchmark datasets unequivocally demonstrate that CoMT-VD achieves the superior generalization performance (in the four integrated scenarios from five benchmark datasets, the improvements were 5.0∼12.0%, 6.0∼12.5%, 5.0∼11.2%, 5.0∼11.2%, and 6.3∼12.3%, respectively), marking a notable advancement towards robust and reliable real-world violence detection systems. Full article

Bimodal emotion recognition based on audio and text is widely adopted in video-constrained real-world applications such as call centers and voice assistants. However, existing systems suffer from limited cross-domain generalization and monolingual bias. To address these limitations, a cross-lingual bimodal emotion recognition method is proposed, integrating Mamba-based temporal encoders for audio (Wav2Vec2.0) and text (Jina-v3) with a Transformer-based cross-modal fusion architecture (BiFormer). Three corpus-adaptive augmentation strategies are introduced: (1) Stacked Data Sampling, in which short utterances are concatenated to stabilize sequence length; (2) Label Smoothing Generation based on Large Language Model, where the Qwen3-4B model is prompted to detect subtle emotional cues missed by annotators, producing soft labels that reflect latent emotional co-occurrences; and (3) Text-to-Utterance Generation, in which emotionally labeled utterances are generated by ChatGPT-5 and synthesized into speech using the DIA-TTS model, enabling controlled creation of affective audio–text pairs without human annotation. BiFormer is trained jointly on the English Multimodal EmotionLines Dataset and the Russian Emotional Speech Dialogs corpus, enabling cross-lingual transfer without parallel data. Experimental results show that the optimal data augmentation strategy is corpus-dependent: Stacked Data Sampling achieves the best performance on short, noisy English utterances, while Label Smoothing Generation based on Large Language Model better captures nuanced emotional expressions in longer Russian utterances. Text-to-Utterance Generation does not yield a measurable gain due to current limitations in expressive speech synthesis. When combined, the two best performing strategies produce complementary improvements, establishing new state-of-the-art performance in both monolingual and cross-lingual settings. Full article

On GPU-based clusters, the training workloads of machine learning (ML) models, particularly neural networks (NNs), are often structured as Directed Acyclic Graphs (DAGs) and typically deployed for parallel execution across heterogeneous GPU resources. Efficient scheduling of these workloads is crucial for optimizing performance metrics such as execution time, under various constraints including GPU heterogeneity, network capacity, and data dependencies. DAG-structured ML workload scheduling could be modeled as a Nonlinear Integer Program (NIP) problem, and is shown to be NP-complete. By leveraging a positive correlation between Scheduling Plan Distance (SPD) and Finish Time Gap (FTG) identified through an empirical study, we propose to develop a Running Time Gap Strategy for scheduling based on Whale Optimization Algorithm (WOA) and Reinforcement Learning, referred to as WORL-RTGS. The proposed method integrates the global search capabilities of WOA with the adaptive decision-making of Double Deep Q-Networks (DDQN). Particularly, we derive a novel function to generate effective scheduling plans using DDQN, enhancing adaptability to complex DAG structures. Comprehensive evaluations on practical ML workload traces collected from Alibaba on simulated GPU-enabled platforms demonstrate that WORL-RTGS significantly improves WOA’s stability for DAG-structured ML workload scheduling and reduces completion time by up to 66.56% compared with five state-of-the-art scheduling algorithms. Full article

Traffic safety is a critical issue in highway operation management, where accurate identification of accident hotspots enables proactive risk prevention and facility optimization. Traditional methods relying on historical statistics often fail to capture macro-level environmental patterns and micro-level dynamic variations. To address this challenge, we propose a Dual-Branch Feature Adaptive Gated Fusion Network (DFAGF-Net) that integrates satellite remote sensing imagery with traffic flow time-series data. The framework consists of three components: the Global Contextual Aggregation Network (GCA-Net) for capturing macro spatial layouts from remote sensing imagery, a Sequential Gated Recurrent Unit Attention Network (Seq-GRUAttNet) for modeling dynamic traffic flow with temporal attention, and a Hybrid Feature Adaptive Module (HFA-Module) for adaptive cross-modal feature fusion. Experimental results demonstrate that the DFAGF-Net achieves superior performance in accident hotspot recognition. Specifically, GCA-Net achieves an accuracy of 84.59% on satellite imagery, while Seq-GRUAttNet achieves an accuracy of 82.51% on traffic flow data. With the incorporation of the HFA-Module, the overall performance is further improved, reaching an accuracy of 90.21% and an F1-score of 0.92, which is significantly better than traditional concatenation or additive fusion methods. Ablation studies confirm the effectiveness of each component, while comparisons with state-of-the-art models demonstrate superior classification accuracy and generalization. Furthermore, model interpretability analysis reveals that curved highway alignments, roadside greenery, and varying traffic conditions across time are major contributors to accident hotspot formation. By accurately locating high-risk segments, DFAGF-Net provides valuable decision support for proactive traffic safety management and targeted infrastructure optimization. Full article

Haze significantly reduces visibility in critical applications such as autonomous driving, surveillance, and firefighting, making its removal essential for safety and reliability. Motivated by the limited robustness of the existing methods under non-uniform haze conditions, this study introduces a novel regression-based dehazing model that simultaneously incorporates the atmospheric light constant, transmission map, and scattering coefficient for improved restoration. Instead of relying on complex deep networks, the model leverages brightness–saturation cues and regression-driven scattering estimation with localized haze detection to reconstruct clearer images efficiently. Evaluated on the RESIDE dataset, the approach consistently surpasses state-of-the-art techniques including Dark Channel Prior, AOD-Net, FFA-Net, and Single U-Net, achieving SSIM = 0.99, PSNR = 22.25 dB, VIF = 1.08, and the lowest processing time of 0.038 s, demonstrating both accuracy and practicality for real-world deployment. Full article

Deaf and hearing-impaired individuals rely on sign language, a visual communication system using hand shapes, facial expressions, and body gestures. Sign languages vary by region. For example, Arabic Sign Language (ArSL) is notably different from American Sign Language (ASL). This project focuses on creating an Arabic Sign Language Recognition (ArSLR) System tailored for healthcare, aiming to bridge communication gaps resulting from a lack of sign-proficient professionals and limited region-specific technological solutions. Our research addresses limitations in sign language recognition systems by introducing a novel framework centered on ResNet50ViT, a hybrid architecture that synergistically combines ResNet50’s robust local feature extraction with the global contextual modeling of Vision Transformers (ViT). We also explored a tailored Vision Transformer variant (SignViT) for Arabic Sign Language as a comparative model. Our main contribution is the ResNet50ViT model, which significantly outperforms existing approaches, specifically targeting the challenges of capturing sequential hand movements, which traditional CNN-based methods struggle with. We utilized an extensive dataset incorporating both static (36 signs) and dynamic (92 signs) medical signs. Through targeted preprocessing techniques and optimization strategies, we achieved significant performance improvements over conventional approaches. In our experiments, the proposed ResNet50-ViT achieved a remarkable 99.86% accuracy on the ArSL dataset, setting a new state-of-the-art, demonstrating the effectiveness of integrating ResNet50’s hierarchical local feature extraction with Vision Transformer’s global contextual modeling. For comparison, a fine-tuned Vision Transformer (SignViT) attained 98.03% accuracy, confirming the strength of transformer-based approaches but underscoring the clear performance gain enabled by our hybrid architecture. We expect that RAFID will help deaf patients communicate better with healthcare providers without needing human interpreters. Full article

Emotion recognition based on electroencephalography (EEG) has gained significant attention due to its potential applications in human–computer interaction, affective computing, and mental health assessment. This study presents a convolutional neural network-based approach to emotion valence prediction model development using 4-channel headband EEG data as well as its evaluation based on computer vision emotion valence recognition. We trained a model on the publicly available FACED and SEED datasets and tested it on a newly collected dataset recorded using a wearable BrainBit headband. The model’s performance is evaluated using both standard train–validation–test splitting and a leave-one-subject-out cross-validation strategy. Additionally, the model is evaluated on computer vision-based emotion recognition system to assess the reliability and consistency of EEG-based emotion prediction. Experimental results demonstrate that the CNN model achieves competitive accuracy in predicting emotion valence from EEG signals, despite the challenges posed by limited channel availability and individual variability. The findings show the usability of compact EEG devices for real-time emotion recognition and their potential integration into adaptive user interfaces and mental health applications. Full article

A Bose-Fermi mixture, consisting of both bosons and fermions, exhibits distinctive quantum coherence and phase transitions, offering valuable insights into many-body quantum systems. The ground state, as the system’s lowest energy configuration, is essential for understanding its overall behavior. In this study, we introduce the Bose-Fermi Energy-based Deep Neural Network (BF-EnDNN), a novel deep learning approach designed to solve the ground-state problem of Bose-Fermi mixtures at zero temperature through energy minimization. This method incorporates three key innovations: point sampling pre-training, a Dynamic Symmetry Layer (DSL), and a Positivity Preserving Layer (PPL). These features significantly improve the network’s accuracy and stability in quantum calculations. Our numerical results show that BF-EnDNN achieves accuracy comparable to traditional finite difference methods, with effective extension to two-dimensional systems. The method demonstrates high precision across various parameters, making it a promising tool for investigating complex quantum systems. Full article

There is pressure on security operations centers (SOCs) from public and private industries as they are coping with the surge of cyberattacks, which is making the reconciliation of inherent organizational tensions a priority. This study surfaces two persistent tensions: (1) expediency versus authority, and (2) adaptability versus consistency that have remained underexplored in cybersecurity literature. We based the research on empirical data collected across three organizational settings, an international consumer packaged goods, a non-departmental public body based in the Netherlands, and a global managed security service provider. Thus, we reveal these not as isolated trade-offs but as paradoxes that must be continuously navigated within SOC operations. Built upon both empirical analysis and Paradox Theory, we develop a conceptual model that explains how SOCs reconcile these tensions through the strategic integration of artificial intelligence (AI), automation, and human expertise. Our model emphases that AI and automation do not replace human analysts; rather, they allow a new form of organizational balance, through mechanisms such as Dynamic Equilibrium and iterative integration. The model demonstrates how SOCs embed technological and human capabilities to sustain simultaneously agility, consistency, authority, and speed. By reframing AI integration as a process of paradox reconciliation, not as a resistance or automation alone, this study contributes new theoretical insight into the sociotechnical dynamics shaping the future of cybersecurity operations. Full article

In recent years, interest in the application of ontologies in various domains of knowledge has grown significantly. Ontologies are widely used in a myriad of areas, such as artificial intelligence, data integration, knowledge management, and the semantic web, to name but a few. However, despite the widespread adoption, there exist a range of problems associated with ontologies, such as the complexity and cognitive challenges associated with ontology engineering, design, and development. One of the solutions to these challenges is to reuse existing ontologies rather than developing new ontologies afresh for new applications. The reuse of ontologies that describe a knowledge domain is a complex task consisting of many aspects. One of the key aspects involves ranking ontologies to aid in their selection. Various techniques have been proposed for this task, but many of them fall short in their expressiveness and ability to capture the cognitive aspects of human-like decision-making processes. Furthermore, much of the existing research focuses on an objective approach to ontology ranking, but it is unquestionable that a wide range of aspects pertaining to the quality of an ontology simply cannot be captured in a quantitative manner. Existing ranking models fail to provide a robust and flexible canvas for facilitating qualitative ontology ranking and selection for reuse. To address the aforementioned shortcomings of existing ontology ranking approaches, this study proposes a novel algorithm for ranking ontologies that extends the Elimination and Choice Translating Reality (ELECTRE) multi-criteria decision-making method with the Linguistic q-Rung Orthopair Fuzzy Set (Lq-ROFS-ELECTRE II), allowing the expression of uncertainty in a more robust and precise manner. The new Lq-ROFS-ELECTRE II algorithm was applied to rank a set of 19 ontologies of the machine learning (ML) domain. The ML ontologies were evaluated using a set of seven qualitative criteria extracted from the Ontometric framework. The proposed Lq-ROFS-ELECTRE II algorithm was then applied to rank the 19 ontologies in light of the seven criteria. The ranking results obtained were compared against the quantitative rankings of the same 19 ontologies using the traditional ELECTRE II algorithm, and confirmed the validity of the ranking performed by the proposed Lq-ROFS-ELECTRE II algorithm and its effectiveness in the task of ontology ranking. Furthermore, a comparative analysis of the proposed Lq-ROFS-ELECTRE II against existing MCDM methods and other existing fuzzy ELECTRE II methods displayed its superior modeling capabilities that allow for more natural decision evaluation from subject experts in real-world applications and allow the decision-maker to have much flexibility in expressing their preferences. These capabilities of the Lq-ROFS-ELECTRE II algorithm make it applicable not only in ontology ranking, but in any domain where there exist decision-making scenarios that comprise multiple conflicting criteria under uncertainty. Full article

Innovative reasoning frameworks have been proposed to enhance the reasoning capabilities of AI agents, improving their performance in various tasks. However, most existing research has focused on enhancing designing frameworks for LLMs, with limited attention on leveraging in-context learning to boost their reasoning power. This paper proposes a novel approach, Demo-ToT, which enhances the Tree-of-Thought (ToT) reasoning framework by dynamically retrieving relevant demonstrations to improve reasoning accuracy. Various demonstration retrieval strategies, including vector similarity, sparse retrieval, and string similarity, were explored to identify the most effective methods for optimizing LLM performance. Experiments conducted across multiple benchmarks and language models of varying sizes demonstrated that Demo-ToT substantially enhanced the reasoning ability of smaller LLMs, achieving performance comparable to or even surpassing that of much larger models such as GPT-4. Full article

In this study, we present LizAI XT, an AI-powered platform designed to automate the structuring, anonymization, and semantic integration of large-scale healthcare data from diverse sources, into one comprehensive table or any designated forms, based on diseases, clinical variables, and/or other defined parameters, beyond the creation of a dashboard or visualization. We evaluate the platform’s performance on a cluster of 4x NVIDIA A30 GPU 24GB, with 16 diseases—from deathly cancer and COPD, to conventional ones—ear infections, including a total 16,000 patients, ∼115,000 medical files, and ∼800 clinical variables. LizAI XT structures data from thousands of files into sets of variables for each disease in one file, achieving > 95.0% overall accuracy, while providing exceptional outputs in complicated cases of cancers (99.1%), COPD (98.89%), and asthma (98.12%), without model-overfitting. Data retrieval is sub-second for a variable per patient with a minimal GPU power, which can significantly be improved on more powerful GPUs. LizAI XT uniquely enables fully client-controlled data, complying with strict data security and privacy regulations per region/nation. Our advances complement the existing EMR/EHR, AWS HealthLake, and Google Vertex AI platforms, for healthcare data management and AI development, with large-scalability and expansion at any levels of HMOs, clinics, pharma, and government. Full article

Medical image segmentation has been a central research focus in deep learning, but methods based on convolutions have limitations in modeling the long-range validity of images. To overcome this issue, hybrid CNN-Transformer architectures have been explored, with SwinUNet being a classic approach. However, SwinUNet still faces challenges such as insufficient modeling of relative position information, limited feature fusion capabilities in skip connections, and the loss of translational invariance caused by Patch Merging. To overcome these limitations, the architecture RE-XswinUnet is presented as a novel solution for medical image segmentation. In our design, relative position biases are replaced with rotary position embedding to enhance the model’s ability to extract detailed information. During the decoding stage, XskipNet is designed to improve cross-scale feature fusion and learning capabilities. Additionally, an SCAR Block downsampling module is incorporated to preserve translational invariance more effectively. The experimental results demonstrate that RE-XswinUnet achieves improvements of 2.65% and 0.95% in Dice coefficients on the Synapse multi-organ and ACDC datasets, respectively, validating its superiority in medical image segmentation tasks. Full article

Multi-hop question answering tasks involve identifying relevant supporting sentences from a given set of documents, which serve as the rationale for deriving answers. Most research in this area consists of two main components: a rationale identification module and a reader module. Since the rationale identification module often relies on retrieval models or supervised learning, annotated rationales are typically essential. This reliance on annotations, however, creates challenges when adapting to open-domain settings. Moreover, when models are trained on annotated rationales, explainable artificial intelligence (XAI) requires clear explanations of how the model arrives at these rationales. Consequently, traditional multi-hop question answering (QA) approaches that depend on annotated rationales are ill-suited for XAI, which demands transparency in the model’s reasoning process. To address this issue, we propose a rationale reasoning framework that can effectively infer rationales and clearly demonstrate the model’s reasoning process, even in open-domain environments without annotations. The proposed model is applicable to various tasks without structural constraints, and experimental results demonstrate its significantly improved rationale reasoning capabilities in multi-hop question answering, relation extraction, and sentence classification tasks. Full article

Big Data Analytics is vital for power grids, as it empowers informed decision-making, anticipates potential operational and maintenance issues, optimizes grid management, supports renewable energy integration, ultimately reduces costs, improves customer service, monitors consumer behavior, and offers new services. This paper describes the AI–Big Data Analytics Architecture based on a data lake architecture that uses a reduced and customized set of Hadoop and Spark as a cost-effective, on-premises alternative for advanced data analytics in power systems. As a case study, a comparative analysis of electricity price forecasting models in the day-ahead market for nodes of the Mexican national electrical system using statistical, machine learning, and deep learning models, is presented. To build and select the best forecasting model, a data science and machine learning methodology is used. The results show that the Gradient Boosting and Support Vector Regression models presented the best performance, with a Mean Absolute Percentage Error (MAPE) between 1% and 4% for five-day-ahead electricity price forecasting. The implementation of the best forecasting model into the Big Data Analytics Platform allows the automation of the calculation of the local electricity price forecast per node (every 24, 72, or 120 h) and its display in a comparative dashboard with actual and forecasted data for decision-making on demand. The proposed architecture is a valuable tool that allows the future implementation of intelligent energy forecasting models in power grids, such as load demand, fuel prices, power generation, and consumption, among others. Full article

Healthcare systems worldwide have faced unprecedented pressure during crises such as the COVID-19 pandemic, exposing limits in managing scarce hospital resources. Many predictive models remain static, unable to adapt to new variants, shifting conditions, or diverse patient populations. This work proposes a dynamic prioritization framework that recalculates severity scores in batch mode when new factors appear and applies them instantly through a streaming pipeline to incoming patients. Unlike approaches focused only on fixed mortality or severity risks, our model integrates dual datasets (survivors and non-survivors) to refine feature selection and weighting, enhancing robustness. Built on a big data infrastructure (Spark/Databricks), it ensures scalability and responsiveness, even with millions of records. Experimental results confirm the effectiveness of this architecture: The artificial neural network (ANN) achieved 98.7% accuracy, with higher precision and recall than traditional models, while random forest and logistic regression also showed strong AUC values. Additional tests, including temporal validation and real-time latency simulation, demonstrated both stability over time and feasibility for deployment in near-real-world conditions. By combining adaptability, robustness, and scalability, the proposed framework offers a methodological contribution to healthcare analytics, supporting fair and effective hospitalization prioritization during pandemics and other public health emergencies. Full article

Network traffic analysis is crucial for understanding network behavior and identifying underlying applications, protocols, and service groups. The increasing complexity of network environments, driven by the evolution of the Internet, poses significant challenges to traditional analytical approaches. Graph Neural Networks (GNNs) have recently garnered considerable attention in network traffic analysis due to their ability to model complex relationships within network flows and between communicating entities. This scoping review systematically surveys major academic databases, employing predefined eligibility criteria to identify and synthesize key research in the field, following the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) methodology. We present a comprehensive overview of a generalized architecture for GNN-based traffic analysis and categorize recent methods into three primary types: node prediction, edge prediction, and graph prediction. We discuss challenges in network traffic analysis, summarize solutions from various methods, and provide practical recommendations for model selection. This review also compiles publicly available datasets and open-source code, serving as valuable resources for further research. Finally, we outline future research directions to advance this field. This work offers an updated understanding of GNN applications in network traffic analysis and provides practical guidance for researchers and practitioners. Full article

Forecasting high-volume, univariate, and multivariate longitudinal data streams is a critical challenge in Big Data systems, especially with constrained computational resources and pronounced data variability. However, existing approaches often neglect multivariate statistical complexity (e.g., covariance, skewness, kurtosis) of multivariate time series or rely on recency-only windowing that discards informative historical fluctuation patterns, limiting robustness under strict resource budgets. This work makes two core contributions to big data forecasting. First, we establish a formal, multi-dimensional framework for quantifying “data bigness” across statistical, computational, and algorithmic complexities, providing a rigorous foundation for analyzing resource-constrained problems. Second, guided by this framework, we extend and validate the Adaptive High-Fluctuation Recursive Segmentation (AHFRS) algorithm for multivariate time series. By incorporating higher-order statistics such as covariance, skewness, and kurtosis, AHFRS improves predictive accuracy under strict computational budgets. We validate the approach in two stages. First, a real-world case study on a univariate Bitcoin time series provides a practical stress test using a Long Short-Term Memory (LSTM) network as a robust baseline. This validation reveals a significant increase in forecasting robustness, with our method reducing the Root Mean Squared Error (RMSE) by more than 76% in a challenging scenario. Second, its generalizability is established on synthetic multivariate data sets in Finance, Retail, and Healthcare using standard statistical models. Across domains, AHFRS consistently outperforms baselines; in our multivariate Finance simulation, RMSE decreases by up to 62.5% in Finance and Mean Absolute Percentage Error (MAPE) drops by more than 10 percentage points in Healthcare. These results demonstrate that the proposed framework and AHFRS advances the theoretical modeling of data complexity and the design of adaptive, resource-efficient forecasting pipelines for real-world, high-volume data ecosystems. Full article