Search Results (490)

Accurate multistep-ahead prediction of natural gas pipeline-network states is essential for intelligent dispatching, yet such networks contain physically heterogeneous components (gas sources, pipelines, compressors, valves), and historical states and future dispatching commands are decoupled in both temporal scale and physical semantics. This causes conventional data-driven models to suffer from semantic entanglement and cumulative error during long-horizon forecasting. This study proposes a deep spatiotemporal surrogate model with three coordinated designs: (i) type-specific feature encoding combined with global latent-graph mapping and a shared graph convolutional network (GCN) to disentangle heterogeneous-equipment attributes and represent network-wide topological coupling; (ii) a residual-gated temporal coupling mechanism that adaptively fuses historical operating inertia with future external disturbances; and (iii) a temporal-gradient multi-objective loss with a 12-step autoregressive rolling strategy over a 6 h horizon to suppress cumulative divergence. On 85,248 samples built from field monitoring data and commercial mechanistic simulations, the model attains median relative errors of 1.15% for nodal pressure and 2.10% for pipeline flow, capturing macroscopic pressure decay and high-frequency transient flow induced by valve and compressor switching without noticeable delay, providing an efficient tool for online simulation, real-time warning, and decision support in complex natural gas pipeline networks. Full article

The rapid expansion of online real estate (RE) platforms has intensified information overload, making property search and decision-making increasingly complex. Real estate recommendation systems (RERSs) have emerged as essential decision-support tools; however, their development has not kept pace with advances in explainable artificial intelligence (XAI), transfer learning (TL), and fairness-aware machine learning. This PRISMA-compliant systematic review synthesizes 59 peer-reviewed studies published between 2005 and 2025 to critically examine algorithmic approaches, data modalities, evaluation practices, and ethical considerations in RERS research. Our analysis reveals a substantial lag in the adoption of state-of-the-art AI techniques: While deep learning is employed in 15% of studies, no reviewed work implements state-of-the-art post hoc XAI or TL frameworks, despite their relevance for addressing interpretability and data scarcity challenges. Furthermore, we identify systemic research biases, including reliance on proprietary datasets (80%), geographic concentration in Asia (56%), the dominance of residential property studies (91%), and limited fairness auditing despite documented discrimination risks in housing markets. To address these gaps, we propose a trust-based evaluation (T-EVAL) framework that integrates predictive accuracy, user trust, fairness, and market efficiency, and introduces a comprehensive nine-layer conceptual architecture for transparent, ethical, and data-efficient next-generation RERS. This review establishes an empirical benchmark for technology adoption gaps and outlines a research agenda for advancing responsible AI in RE decision-support systems. Full article

In this work, we present a neural network-based study capable of predicting faults in oil-insulated power transformers through the analysis of dissolved gases. The advantage of this study lies in using data already collected by electric power companies, which gather it to comply with international or regional standards; however, they sometimes act only after the equipment is already in a faulty condition. Therefore, the challenge in this work was data regularization, as collections typically occur at long intervals of 6 to 12 months. Furthermore, samples are often irregular, as data collection depends on factors such as weather and the availability of maintenance teams. As a result of this work, Multilayer Perceptron (MLP), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM) were used to predict failures with advanced forecasts ranging from 1 to 6 months, achieving accuracies of 97.5% and 85%, respectively. Thus, these models prove to be important tools for maintenance planning, enabling adequate predictability for organizing equipment shutdowns without the need for high investments in installing tools to capture this information online and adapting substations to send data to control rooms or other analysis centers. Full article

Current deep learning methods for tool wear monitoring suffer from poor interpretability and struggle to reveal the intrinsic relationship between signals and wear states. To address this issue, this paper presents an interpretable tool wear monitoring method based on Sparse Identification of Nonlinear Dynamics (SINDy). Multi-domain features are extracted from cutting force and acoustic emission signals to construct a time series. The SINDy algorithm is used to identify ordinary differential equations that describe the evolution of tool wear. An iterative “predict-validate-correct” mechanism is applied to optimize model parameters. Experimental results show that the mean absolute percentage error (MAPE) between the predicted and actual values is below 6%. Moreover, the optimal model demonstrates an average MAPE as low as 0.067% in cross-condition tests. This study provides an effective solution for online tool wear monitoring that achieves high precision, strong generalization, and physical interpretability. Full article

Mechanistic interpretability seeks to reverse-engineer the computational circuits within large language models, but current methods rely on exhaustive or heuristic search over exponentially many feature interactions. This paper introduces Active Circuit Discovery (ACD), a framework that combines attribution-graph analysis with active inference to select interventions efficiently. ACD uses Anthropic’s circuit-tracer library as its attributiongraph backend, applying Edge Attribution Patching with transcoders to identify the active transcoder features for each prompt. A partially observable Markov decision process (POMDP) agent, implemented with pymdp, maintains a multi-factor generative model of feature importance, layer role, and causal influence. At each step, the agent selects both a target feature and an intervention type (ablation, activation patching, or feature steering) by minimising Expected Free Energy over the joint feature–action space, and it learns its observation model online through Dirichlet parameter updates. ACD is an interventionselection layer over existing attribution-graph tools; it is not a whole-circuit discovery method, and no claim of state-of-the-art circuit discovery is made. The framework is evaluated on Gemma-2-2B (26 layers) and Llama-3.2-1B (16 layers) across four settings: Indirect Object Identification (IOI), multi-step reasoning, feature steering, and a multidomain benchmark spanning geography, mathematics, science, logic, and history. With a budget of 20 interventions per prompt, an ablation-only agent scored by bounded oracle efficiency against the ablation oracle reaches 82.0% efficiency on Gemma IOI and 73.0% on Gemma multi-step. It exceeds random selection by 43.5% (relative) on Gemma IOI (paired permutation p = 0.031) and is competitive with greedy ranking, a heuristic UCB bandit, and a plain UCB baseline. A direct Edge-Attribution-Patching ranking is itself a strong baseline that the agent does not consistently surpass, and on Llama multi-step the agent reaches 9.3% efficiency (37.8% with finer layer-role bins). All comparisons report bootstrap 95% confidence intervals. The full multi-action agent is characterised separately by a Relative Cumulative KL, a steering-driven amplification factor reported apart from the bounded efficiency. Feature steering changes the top-1 prediction in a dose-dependent manner, but a matched random-feature control shows that circuit-selected features are only marginally, and not significantly, more steerable than random active features at large multipliers, indicating that part of the effect is generic activation scaling. Multi-domain analysis shows task-dependent circuit structure, with IOI circuits concentrated in late layers and reasoning and scientific knowledge recruiting early and middle layers. Code, notebooks (free T4), AMD64/aarch64 Docker images, and raw results are publicly available. Full article

Background/Objectives: Helplines are an important component of mental health support systems; however, limited research has examined burnout among helpline workers. This exploratory pilot study investigated the relationships between job demands, job and personal resources, and burnout within the Job Demands–Resources (JD-R) framework. Methods: A cross-sectional online study was conducted among 73 helpline workers. Burnout was assessed using the Burnout Assessment Tool (BAT). Job demands were categorized as topic-related, client-related, and service-related. Data were analysed using Spearman correlations and hierarchical regression analyses. Results: Job demands were positively associated with burnout (r_s = 0.41, p < 0.001), while job resources were negatively associated with burnout (r_s = −0.30, p < 0.01). Regression analyses showed that job demands and resources explained 50% of the variance in overall burnout (R² = 0.50, p < 0.001). Service-related demands emerged as the strongest predictor of burnout (β = 0.70, p < 0.001) and consistently predicted exhaustion (β = 0.60, p < 0.001), mental distance (β = 0.48, p < 0.001), cognitive impairment (β = 0.52, p < 0.001), and emotional impairment (β = 0.61, p < 0.001). Organizational resources were negatively associated with mental distance (β = −0.31, p < 0.01), whereas topic-related demands were not significant predictors. Conclusions: The findings highlight the importance of differentiating types of job demands in understanding burnout among helpline workers. Service-related demands appeared to be more strongly associated with burnout than topic- or client-related demands, suggesting that structural aspects of helpline work may be particularly relevant for worker well-being. Full article

This study examines the systemic structures that limit inclusivity, diversity, equality, and accessibility (IDEA) in South Africa’s construction industry. It develops an empirically grounded framework, linking digital platform/tool (software tools and systems that facilitate construction processes) adoption to the institutional changes needed to advance gender equity. Building on a literature review, an online survey of 112 Construction Industry Development Board (Cidb)-registered practitioners was analyzed in SPSS v26 using descriptive and inferential statistics and principal component analysis (PCA). Results show that gender differences in mastery of core digital tools were not statistically significant (p > 0.05 across tool categories). The regression model predicting perceived career growth showed weak explanatory power and was not statistically significant (R² = 0.068; F(10,100) = 0.734; p = 0.691). Accordingly, the non-significant model is interpreted as indicating that the predictors included are insufficient to explain perceived career growth in this sample, and that other organizational and structural factors may be more influential. PCA produced a three-component digital inclusivity ecosystem composed of operational fairness, technical empowerment, and integrative leadership, demonstrating 62.84% variance explained and a three-pillar systemic architecture for equity composed of legislative frameworks, socioeconomic support, and organizational practice. Leadership representation remained skewed (73.83% male overall; 78.64% at the director level). The study concludes that progress toward IDEA is more likely to result from combining digital adoption with multi-level institutional reforms. Practical implications include integrated policy interventions and organizational practices that address structural barriers while leveraging digital platforms for inclusion. Full article

Online grocery retailing remains effort-intensive because consumers must coordinate many item-level decisions, making basket construction and correction a key barrier to more frequent use. This study examines consumer acceptance of an AI-pre-filled basket in Austrian online grocery retailing and links Technology Acceptance Model (TAM) mechanisms to objective interaction performance. Using a field-proximate one-group pre–post design, participants edited an AI-pre-filled basket in a standardized online shop and completed pre- and post-task surveys using 0–100 slider scales; log data captured processing time and edit actions. Analyses are based on n = 297 valid cases. Results show a substantial and statistically significant within-person increase in online grocery usage intention after the AI-basket interaction, rising from 39.17 to 59.78. The TAM results support the core mechanism: perceived ease of use is positively associated with perceived usefulness, and perceived usefulness strongly predicts behavioral intention, whereas a direct ease-of-use effect on intention is not supported. Linking perceptions to log data shows that acceptance is more strongly associated with correction demand than with processing time. AI-basket acceptance depends less on speed than on reducing rework while preserving user control; retailers should therefore design AI baskets as controllable, editable, low-rework systems rather than speed-oriented automation tools. Full article

The steady advancement of online education has not automatically translated into improved educational quality. Teacher training often continues to focus on the technical use of digital tools, while the pedagogical processes through which teachers report supporting students’ digital competence remain insufficiently understood. The objective of this study was to examine the sequential and predictive structure of teachers’ digital competence using the DigCompEdu framework as a reference. A quantitative cross-sectional study was conducted with a sample of 136 university teachers involved in online education. Data were collected through a self-reported questionnaire based on DigCompEdu and analyzed in two phases: Partial Least Squares Structural Equation Modeling (PLS-SEM) and Artificial Neural Networks (ANNs). The PLS-SEM results suggested a sequential pattern of associations among teacher-reported constructs: Professional Commitment (PC) was positively associated with Digital Resource Management (DR), which in turn was positively associated with Digital Pedagogy (DP) and Assessment and Feedback (AF). These dimensions were associated with Student Empowerment (SE), which showed the strongest positive relationship with teachers’ reported practices for Facilitating Students’ Digital Competence (FS). The ANN sensitivity analysis showed adequate predictive performance in the testing phase (RMSE = 0.155) and identified Student Empowerment as the predictor with the highest normalized importance within the specified model. These findings suggest that faculty development in online higher education may benefit from moving beyond basic digital literacy and platform management toward pedagogical design, formative assessment, inclusive participation, and learner agency. However, the results should be interpreted as evidence of teacher-reported facilitation practices within the analyzed sample, rather than as direct evidence of students’ actual digital competence development. Full article

Background/Objectives: Parenting interventions are an effective way to support child development, and brief screening tools can support equitable implementation of parenting interventions by reducing program costs, increasing accessibility, and engaging populations who have traditionally been underserved. However, brief assessments are frequently overlooked and underutilized. The Family Check-Up (FCU) Online is a digital parenting intervention that integrates a brief FCU Online Assessment, feedback, and parenting skills via an app along with optional provider support. To date, no prior work has validated the FCU Online Assessment. Method: The current study combined two samples of parents participating in FCU Online studies and assessed: (1) reliability, (2) construct validity, (3) convergent validity by comparing FCU Online Assessment subscales to similar parenting and child behavior measures, and (4) predictive validity by using FCU Online Assessment at pretest to predict posttest scores as well as parenting and child behaviors at time 2 and time 3. Results: Strong reliability was found among all five subscales, including Low Conflict (7 items, α = .81), Positive Parenting Practices (11 items, α = .80), Positive School Behaviors (5 items, α = .83), Consistent Rules and Routines (11 items, α = .81), and Child Mental Health (5 items, α = .80). The FCU Online Assessment demonstrated construct and convergent validity, as well as predictive validity in that the FCU Online Assessment at pretest predicted posttest scores. Conclusions: The FCU Online Assessment is a brief, reliable, and valid measure of parenting and child wellbeing. It can be used by parents and providers alike to evaluate parenting skills and child mental health, develop targeted goals and intervention approaches, and assess family wellbeing over time. Full article

This paper presents a comprehensive review on near-infrared (NIR) spectroscopy applied in food analysis, systematically elaborating its core principles, widespread industrial applications, advanced chemometric strategies, and cutting-edge technological progress. NIR spectroscopy (760–2500 nm), characterized by rapid, non-destructive detection and minimal sample preparation, has been widely implemented in quality evaluation and safety monitoring of grains, meat, fruits and vegetables, dairy, fermented products, tea, coffee, and other processed foods, realizing quantitative analysis of nutrients, freshness assessment, texture prediction, adulteration identification, origin tracing, and rapid preliminary screening of toxin/pesticide residues. A series of chemometric methods, including spectral preprocessing (SNV, MSC, S-G smoothing), feature extraction, and variable selection (CARS, PSO-CMW, ICPA), as well as linear/nonlinear modeling algorithms (PLS, SVM, BP-ANN, fuzzy clustering) significantly boost the accuracy and robustness of spectral analysis. Meanwhile, portable NIR devices and online monitoring systems promote on-site and real-time detection in food supply chains. Despite existing challenges such as calibration transfer, matrix interference, and model generalization, innovations like multimodal data fusion, deep learning integration, and intelligent algorithm optimization offer effective solutions. This review not only summarizes the latest research advances of NIR technology in the food field but also emphasizes its significant advantages as a rapid, non-destructive complementary tool to traditional destructive detection methods, providing theoretical support and technical reference for accelerating the industrial translation and standardized application of NIR spectroscopy, and ultimately safeguarding global food quality and safety. Full article

Background/Objectives: Several subunit vaccines for tuberculosis (TB), such as MVA85A and H4:IC31, have not demonstrated ideal protective efficacy in clinical trials, which may be attributed to their limited antigenic profile and lack of effective latency-associated antigens. In this study, we combined bioinformatics with experimental validation to screen for latency-associated antigens that have immune-protective effects. Methods: Highly expressed antigens were identified from models related to latent infections, such as hypoxia and nutritional starvation. Their physicochemical properties and immunogenicity were predicted using online tools such as Expasy-ProParam, IEBD, and VaxiJen. The immunogenicity of these antigens was then evaluated in multiple mycobacterium infection models. Finally, a systematic evaluation of the immune response and protective effects induced by the candidate antigens was performed in a mouse model using intracellular cytokine detection, mycobacterium growth inhibition assays (MGIAs), antibody-dependent cellular phagocytosis (ADCP), and a latent tuberculosis infection (LTBI) mouse model. Results: The antigen Rv2656c is highly expressed in the nutritional starvation model and demonstrates strong immunogenicity in both infected humans and cattle. Moreover, Rv2656c exerted a significant inhibitory effect against Mycobacterium tuberculosis (M. tuberculosis) and Mycobacterium avium (M. avium) infections in MGIA. The humoral immune response elicited by Rv2656c enhanced the phagocytosis and killing of Mycobacteria by macrophages in vitro. Furthermore, in a mouse model of LTBI established using the attenuated M. tuberculosis H37Ra strain, treatment with Rv2656c significantly decreased the bacterial load in the lungs of the mice. Conclusions: Latency-associated Rv2656c may serve as an immune-protective antigen, offering potential for the development of novel multi-stage antigen subunit vaccine against TB. Full article

This study aimed to investigate pre-service teachers’ perceptions of coding, including their acceptance of coding and their experiences of using a block-based tool to code. With a focus on acceptance and affective factors, we examined the influence of proposed factors on pre-service teachers’ intention to adopt coding, as well as the relationships of acceptance factors. Participants were pre-service teachers from a university in the northeastern United States. Data were collected using an online survey. Both quantitative and qualitative approaches were performed to analyze the data. The results indicated that pre-service teachers’ affective experiences significantly influenced their intention to code. Pre-service teachers’ perceived ease of use and usefulness significantly predicted their intention to code. Gender and coding skills played an important role in pre-service teachers’ intentions for coding adoption. Overall, pre-service teachers’ experiences with the coding activity were positive, leading to more positive changes in their perceptions of coding than negative changes. Full article

The research examines regional disparities in the diffusion of e-commerce among enterprises employing at least 10 people in Italy, using an integrated analytical framework that blends econometric modeling, machine learning, and network analysis. Instrumental Variable (IV) panel models overcome endogeneity arising from digital infrastructure, socioeconomic factors, and online business activity, with geographic slope as a suitable instrument for broadband penetration. Machine learning models—regularized regressions, random forests, and boosting—augment causal inference by registering nonlinear effects and sorting variable salience. The results, in all cases, emphasize internet use, household digital connectivity, and the prevalence of remote work as the most important predictors of the diffusion of e-commerce. Cluster analysis identifies regional digital profiles that distinguish northern-central regions from southern-insular regions, characterizing persistently distinct digital divides. The network analysis, in turn, identifies digital inclusion variables—such as internet penetration and ICT infrastructure—that occupy central positions within the entirety of the economic and technological interdependencies’ regime. Innovation and income levels, while practiced, hold peripheral positions, indicating that digital capacity, rather than economic affluence in the singular, drives online business participation. Italy’s case can particularly illustrate this beyond its national borders. Being a high-income economy with significant regional disparities, it reproduces challenges common elsewhere in the world, among which the cases of Spain, Germany, the USA, the Republic of Korea, and Japan come to mind, where regional disparities inhibit inclusive digital development. The Italian case presents, then, a transferable model for the diffusion of digital tools, the reduction in regional disparities, and the encouragement of economic integration. By synthesizing the causal, predictive, and systemic methodologies, the study offers a theoretical and practical response to digital transformation across diverse terrains. Full article

In high-value precision machining, existing tool wear monitoring models often suffer from two major limitations: poor generalization under varying cutting conditions and heavy reliance on large amounts of labeled data for new operating scenarios. These limitations hinder the practical deployment of intelligent monitoring systems. To address these challenges, this paper proposes a few-shot warm-start framework based on model-agnostic meta-learning. The method consists of two stages. First, meta-training is performed on historical machining data to learn a task-sensitive parameter initialization that enables rapid adaptation. Second, under a new operating condition, the few-shot warm-start mechanism collects a minimal number (1 to 5) of samples through a targeted physical trial-cutting process for online fine-tuning, aligning the model with the current physical environment. Experiments on the PHM2010 dataset fully simulate varying cutting scenarios. The experimental results demonstrate that the proposed framework consistently outperforms traditional transfer learning, deep learning models, and existing meta-learning approaches, offering an effective solution for fast and accurate tool wear prediction under few-shot and varying cutting conditions. Full article