This research paper investigates the integration of Explainable Artificial Intelligence (XAI) into Predictive Maintenance (PdM) systems, aiming to enhance transparency, interpretability, and reliability in industrial applications. The primary contribution is the introduction of the Explainability Parameters (XPA) framework, which offers a structured methodology for evaluating and applying XAI in PdM. The study systematically reviews recent advancements and challenges in the literature, categorising explanations into pre-modelling, in-modelling, and post-modelling processes. It presents and analyses significant case studies across various industrial sectors to illustrate the practical implications and hurdles of XAI methodologies. Key findings indicate that while XAI significantly improves the effectiveness and trustworthiness of PdM by clarifying model predictions, its implementation is hindered by the complexity of industrial data and the absence of standardised evaluation methods. The XPA framework addresses these challenges by providing tailored metrics for specific applications and advocating for a multi-phase approach to convert technical outputs into actionable maintenance recommendations. The originality of this paper lies in its comprehensive review and the establishment of rigorous standards for assessing XAI methodologies, thereby bridging the gap between theoretical frameworks and practical applications. By promoting adaptable XAI frameworks that cater to real-world industrial needs, this study fosters trust in automated decision-making processes. It enhances the overall understanding of XAI's role in PdM.

This study investigates whether annual report readability influences the project cost of equity capital (COE) in the emerging market of Egypt. To reassess this relationship within a project evaluation context, the research develops a novel, explainable heterogeneous ensemble model capable of capturing complex nonlinear interactions among financial and textual determinants affecting COE estimation. The proposed ensemble integrates Gradient Boosting Regression (GBR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Elastic Net using an average-voting strategy. Across multiple evaluation metrics, the ensemble model outperforms linear regression benchmarks, homogeneous bagging and boosting ensembles, and its individual base learners. Specifically, the model achieves superior predictive performance, with an R² of 0.2538, Mean Squared Error (MSE) of 0.0078, Mean Absolute Error (MAE) of 0.0656, and Root Mean Squared Error (RMSE) of 0.0885, significantly outperforming both linear regression and state-of-the-art machine learning alternatives. Feature importance analysis using RF shows that the market-to-book ratio (MTB) and return on equity (ROE) contribute most to predictive accuracy (18.1% and 18.7%, respectively), highlighting the dominant role of financial fundamentals in project COE estimation. Conversely, readability measures exhibit minimal influence. Shapley Additive Explanations (SHAP) further confirm that annual report readability does not exert a statistically meaningful impact on COE within the Egyptian context. By leveraging advanced machine learning and explainability techniques, this study enhances understanding of COE determinants and offers evidence-based insights to improve project appraisal, financial planning, and strategic decision-making.

The increasing complexity and frequency of cyberattacks have made accurate and reliable intrusion detection systems (IDSs) essential for modern network security. In this study, an explainable triple-hybrid deep learning framework is proposed for binary intrusion detection using the CICIDS2017 dataset. The proposed architecture integrates three complementary branches, namely a Transformer encoder, a bidirectional long short-term memory (BiLSTM) network, and a multilayer perceptron (MLP), to capture global feature interactions, sequential dependencies, and nonlinear discriminative patterns from network traffic data. To enhance adaptive representation learning, the framework employs a branch-gating mechanism and a fusion-gating module before final classification. The model was evaluated in a Benign-versus-Attack setting using 5-fold stratified cross-validation and assessed through accuracy, precision, recall, F1-score, ROC-AUC, confusion matrices, and explainability analysis. Experimental results showed strong and stable performance across folds, with a mean validation accuracy of 97.18%, a best-fold accuracy of 97.43%, and a mean ROC-AUC of 0.9975. LIME-based explanations further improved transparency, confirming the framework as an effective and interpretable solution for binary intrusion detection.

The recent explosion of cyber threats in big data ecosystems has exposed the vulnerability of black-box machine learning models that prioritize accuracy over explainability. Traditional cybersecurity mechanisms do not create a sense as to why a prediction is provided and the analysts are left with doubts and the response mechanisms are slow. This is not very transparent and therefore compromises the operational trust and the traceability of high-risk decisions in real-time defense infrastructures. Current explainable AI (XAI) methods, despite their usefulness, are mostly stagnant, not connected to the changing situation of network behavior and human monitoring. They seldom inculcate the feedback mechanisms that can adjust the model explanations with new threat patterns. The research paper introduces an Explainable Artificial Intelligence and Data-Driven Security Analytics Framework that is hybrid and serves to combine global interpretability with SHAP, local reasoning with LIME, and adaptive refinement with an analyst-in-loop feedback layer. The architecture converts the threat detection to a self-fixing transparent process in which the analytical reasoning is dynamically developed with the data flow. In experimental tests of less than 10k concurrent traffic conditions, experimental results indicate that the system had a detection accuracy of 98.4 and Spearman correlation (0.91) between predicted risk scores and real levels of severity with a quantifiable 3.9% stability improvement over the similar XAI-based systems of intrusion detection. The combination of explainability with real-time analytics will further enhance the accuracy of detecting cyber threats and its interpretability reliability as well. The findings indicate the essential change to credible, self-explanatory, and adaptive security intelligence of the next-generation data networks.

Wearable and Internet of Medical Things devices increasingly support continuous health monitoring, but cloud-dependent analytics remain constrained by latency, connectivity, privacy, and battery requirements. Tiny machine learning shifts inference toward resource-constrained microcontrollers and edge processors; however, its clinical value depends not only on predictive accuracy but also on energy efficiency, real-time responsiveness, and understandable decision logic. This scoping review mapped the evidence on explainability and energy efficiency in TinyML-based health monitoring. Following the Joanna Briggs Institute approach and PRISMA guidance for scoping reviews, Scopus was searched in the title, abstract, and keyword fields for studies published from 2020 to 2025. The search combined TinyML and embedded or edge artificial intelligence terms with wearable or Internet of Medical Things concepts, health-monitoring applications, and explainability or efficiency terms. The supplied export contained 293 records. After screening, 39 reports underwent eligibility assessment and 36 studies were included. Publication activity accelerated sharply, with 18 studies published in 2025. The evidence covered cardiac monitoring, human activity and fall detection, neurological and affective assessment, respiratory monitoring, signal-quality control, glucose sensing, gait analysis, and smart textiles. Convolutional neural networks and hybrid deep models were common, while deployment platforms ranged from microcontrollers to field-programmable gate arrays, application-specific integrated circuits, and neuromorphic hardware. Quantization, pruning, binary or ternary computation, feature reduction, event-driven processing, and local transmission control were frequently used to reduce resource demand. In contrast, only a small minority of studies explicitly evaluated explainability through model-based feature selection, feature importance, or class activation maps. The field is therefore energy-aware but not yet consistently explanation-aware. Future research should adopt standardized hardware reporting, clinician-centered explanation evaluation, external and longitudinal validation, and multiobjective optimization that jointly considers clinical accuracy, energy, latency, memory, robustness, and interpretability.