Accurate prediction of traffic accident severity remains challenging due to feature coupling and class imbalance, which hinder reliable applications in autonomous driving safety systems. This study proposes a Dynamic and Static Cross Entropy Integrated Neural Network (DSCE-INN) to address these issues. Using 857 real-world accident cases from the 2017–2021 China National Automobile Accident In-Depth Investigation System (NAIS), a Weighted Injury Coefficient is developed to enable continuous injury mapping, and K-means clustering reclassifies severity into three levels: property damage only, non-disabling injury, and disabling or fatal injury. Information gain identifies 11 critical features. DSCE-INN employs feature decoupling, transforming the multi-class task into binary sub-models, and introduces a dynamic-static weighted cross-entropy loss to jointly mitigate coupling and imbalance. A soft-hard voting mechanism, combined with L1 regularisation and focal loss, further enhances prediction robustness. Experimental results show accuracies of 0.782, 0.729, and 0.801, significantly outperforming a baseline ANN. Findings demonstrate DSCE-INN’s effectiveness and practical value for autonomous driving safety.

Enhancing maritime accident causation analysis through a hybrid machine learning approach

Linking landscape features of urban parks to visitors' sense of place: A novel hybrid machine learning approach

Optimizing monthly solar PV tilt angles and energy yield across global climate zones: A hybrid machine learning and PVLib approach

A novel hybrid machine learning approach for biorefinery products in pesticide-rich wastewater

A hybrid machine learning approach for customer valuation and decision-making in peer-to-peer lending

Hybrid Numerical and Machine Learning Approaches for Solving Einstein Constraint Equation

Hybrid Machine Learning Approach for Weather Pattern Recognition and Anomaly Detection Using Self-Organizing Maps and K-Nearest Neighbours

Red teaming is a common cybersecurity practice that simulates real-world adversarial cyber operations on defended systems to identify vulnerabilities. Current red team tools often have limited logging capabilities, resulting in insufficient analysis that prevents red teams from receiving real-time feedback and insights after operations. The application of machine learning for automating the analysis of red team operations is severely constrained by the scarcity of labeled, real-world log data. This research addresses this challenge by exploring the potential of using synthetic data to train attack-detection models for Cobalt Strike logs. We systematically evaluate three different training approaches for analyzing Cobalt Strike operational logs: synthetic-only, real-world-only, and a hybrid approach that combines both data types. Our methodology employs a comprehensive feature engineering pipeline that includes both programmatic log generation for creating large-scale structured data and large language model techniques for introducing variety and edge cases. We transform each log file into a high-dimensional vector that includes event types, command verbs, temporal activity patterns, and mappings to the MITRE ATT&CK knowledge base. Random Forest classification models are trained using this feature set to distinguish between successful and failed attack scenarios. By rigorously testing each training approach against a manually labeled ground-truth set of 112 authentic Cobalt Strike logs, we quantify the performance and limitations of each strategy. Our main contribution is demonstrating that a hybrid training strategy achieves 94% accuracy, greatly surpassing synthetic-only models (56%) and real-world-only models (79%). This combined approach effectively addresses both the domain gap in synthetic data and the data scarcity in small, real-world datasets. The hybrid model learns attack diversity from over 30,000 synthetic scenarios while grounding understanding in the authentic structural patterns of real logs, providing a 15 percentage-point improvement over real-data-only approaches. This research offers a practical framework for enhancing limited real-world cybersecurity datasets by strategically integrating synthetic data, enabling immediate use in Department of Defense red team operations and wider cybersecurity machine learning applications.

Eco-friendly concrete mix design incorporating steel slag, textile sludge, and polypropylene fibers through a hybrid experimental and machine learning approach

Abstract Energy audits in wastewater treatment plants (WWTPs), conducted under ISO 50002 and aligned with ISO 50001, are essential to improve sustainability in a highly energy-intensive sector. However, the nonlinearity of biological processes, operational variability, and the limited number of conventional indicators hinder the accurate identification and interpretation of energy inefficiencies. This study proposes SEA-WWTPs, a hybrid quantitative methodology combining deterministic diagnostics with machine learning models trained on event-level features derived from monitored energy performance indicators (EnPIs), to detect and classify energy-inefficiency events and rank likely root causes to support audit-oriented corrective decision-making. A case study in a leachate wastewater treatment plant (WWTP-L) in Portugal analyzed 484,810 1-min records collected over approximately three years. A total of 28,608 inefficiency occurrences (aggregated events) were detected; 96.25% were persistent, indicating predominance of structural and electromechanical causes. The random forest model achieved a Macro-F1 of 0.716 and a Top-2 accuracy of 0.788 using equipment-grouped GroupKFold validation. Overall, the proposed methodology enhances energy audits by integrating data-driven decision support, ensuring traceability, audit-oriented interpretability, and compatibility with ISO standards, thereby supporting the digitalization of sustainable energy management in WWTPs and contributing to the sustainable development goals (SDGs).

Transmission lines are vital components of electrical grids, ensuring the efficient transfer of electricity from power plants to consumers over extensive geographical areas. These lines are constructed with careful consideration of factors such as conductor materials, insulation levels, current ratings, and voltage ratings to maintain reliable and safe electricity delivery. However, various types of faults can occur in transmission lines, posing significant challenges, often leading to outages, equipment damage, and reduced system reliability. Accurate and fast fault classification is therefore a pressing requirement in modern smart grids, where proactive maintenance and resilience are critical. This research addresses the critical need for an efficient electric fault classification model. A comprehensive investigation is conducted, employing a variety of machine learning (ML) algorithms, including Decision Tree (DT), Random Forests (RF), Naive Bayes (NB), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and, AdaBoost, for fault classification. Additionally, fundamental ensemble techniques such as Hard-Voting, Soft-Voting, Stacking, and Blending are incorporated with five hybrid ML models (each constructed by combining various ML algorithms) to enhance fault classification performance and the reliability of transmission lines. Also, this research proposes a hybrid ML model, specifically (RF + DT + Stacking), to classify transmission line data. The main contribution of this work is an application-oriented evaluation of classical and ensemble machine learning models for electrical fault classification, with an emphasis on benchmarking performance, model interpretability, and computational efficiency. This study demonstrates that a carefully configured hybrid ensemble (RF + DT + Stacking) can provide a practical and lightweight alternative to deep learning-based methods in grid fault monitoring scenarios. The dataset used encompasses various attributes affecting line performance, making accurate classification critical for proactive issue detection, optimized maintenance scheduling, and uninterrupted energy supply. Our hybrid model achieves high-performance metrics, including an accuracy of 93.64%, precision of 93.65%, recall of 93.64%, and F1 score of 93.64%, underscoring its effectiveness in enhancing decision-making processes and operational efficiency within electrical transmission networks.

Transmission lines are vital components of electrical grids, ensuring the efficient transfer of electricity from power plants to consumers over extensive geographical areas. These lines are constructed with careful consideration of factors such as conductor materials, insulation levels, current ratings, and voltage ratings to maintain reliable and safe electricity delivery. However, various types of faults can occur in transmission lines, posing significant challenges, often leading to outages, equipment damage, and reduced system reliability. Accurate and fast fault classification is therefore a pressing requirement in modern smart grids, where proactive maintenance and resilience are critical. This research addresses the critical need for an efficient electric fault classification model. A comprehensive investigation is conducted, employing a variety of machine learning (ML) algorithms, including Decision Tree (DT), Random Forests (RF), Naive Bayes (NB), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and, AdaBoost, for fault classification. Additionally, fundamental ensemble techniques such as Hard-Voting, Soft-Voting, Stacking, and Blending are incorporated with five hybrid ML models (each constructed by combining various ML algorithms) to enhance fault classification performance and the reliability of transmission lines. Also, this research proposes a hybrid ML model, specifically (RF + DT + Stacking), to classify transmission line data. The main contribution of this work is an application-oriented evaluation of classical and ensemble machine learning models for electrical fault classification, with an emphasis on benchmarking performance, model interpretability, and computational efficiency. This study demonstrates that a carefully configured hybrid ensemble (RF + DT + Stacking) can provide a practical and lightweight alternative to deep learning-based methods in grid fault monitoring scenarios. The dataset used encompasses various attributes affecting line performance, making accurate classification critical for proactive issue detection, optimized maintenance scheduling, and uninterrupted energy supply. Our hybrid model achieves high-performance metrics, including an accuracy of 93.64%, precision of 93.65%, recall of 93.64%, and F1 score of 93.64%, underscoring its effectiveness in enhancing decision-making processes and operational efficiency within electrical transmission networks.

Exposure to respirable and inhalable dust from engineered stone is linked to lung diseases such as silicosis and COPD, yet the physicochemical properties affecting epithelial-to-mesenchymal transition (EMT) remain unclear. Here, 41 physicochemical properties were characterized across 30 dust samples and evaluated for associations with EMT progression in A549 lung epithelial cells after 24-h exposure. EMT was assessed using three hallmarks: E-cadherin downregulation, vimentin upregulation, and α-SMA upregulation. A hybrid feature selection strategy combining correlation filtering with LassoLarsCV reduced feature redundancy and improved model robustness. The selected features were modeled using optimized regressors (Extreme Gradient Boosting regressor for E-cadherin and Vimentin; Support Vector Machine for α-SMA), and SHAP analysis quantified each property's contribution. Crystalline silica emerged as the most influential factor, showing negative associations with E-cadherin and positive associations with Vimentin and α-SMA. In contrast, sodium-, aluminum-, and rutile-bearing components were associated with lower EMT progression, likely reflecting their occurrence within less reactive mineral phases than crystalline silica. Specific surface area and absolute ζ potential were positively associated with the EMT, indicating enhanced particle-cell interactions and surface-related signaling. These findings establish a framework for linking dust physicochemical characteristics to marker-specific EMT responses and demonstrate the effectiveness of interpretable machine learning for particulate toxicity assessment.

Spatio-Temporal Prediction of Groundwater Level and Quality using IoT Sensing and Hybrid Machine Learning Approaches

Purpose: The complexity of drilling activities has been enhanced by deeper wells, the heterogeneous formations, and the need to provide cost-effective and time-saving hydrocarbon production. One of the most important parameters of drilling performance is rate of Penetration (ROP), which has a direct impact on the efficiency of operations, non-productive time (NPT), and costs. The traditional mechanistic and empirical ROP models that had been important in the past are not very useful in nonlinear interaction, dynamic drilling conditions, and heterogeneous lithologies. However, existing reviews lack a structured problem statement that clearly identifies the limitations of standalone ML and classical ROP models under dynamic drilling conditions and the need for hybrid frameworks that improve accuracy, robustness, and real-time applicability. This review addresses this gap by systematically analyzing hybrid ML approaches and their role in drilling optimization. Materials and Methods: Improved drilling optimization through machine learning (ML) methods, especially hybrid ML models, has redefined the future of drilling optimization, which unites the advantages of various predictive models to improve accuracy, strength, and generalization. This review is a synthesis of literature on hybrid ML applications in ROP prediction, which is divided into three categories: optimization-integrated, ensemble, soft computing, and physics-informed models. Their methodologies, data requirements, real-time integration, operational problems, and performance in comparison to standalone ML models are addressed in the paper. Findings: The essential restrictions, including data quality, computing aspects, and the problem of interpretability, are identified, and the future research direction is also outlined. The synthesis offers an organized scheme of comprehending the development of hybrid ML models in the drilling optimization and outlines opportunities of future progress within the limitations of technologies. Unique contribution to theory, practice and policy: Improved drilling optimization through machine learning (ML) methods, especially hybrid ML models, has redefined the future of drilling optimization, which unites the advantages of various predictive models to improve accuracy, strength, and generalization.

The use of recycled materials as internal curing (IC) agents offers substantial benefits to the concrete industry by improving performance and enhancing environmental sustainability. However, the design of IC concrete has grown intricate due to the nonlinear interactions among many input variables. Previous research on IC is mostly experimental, with only a few studies focusing on predicting the compressive strength (CS) of IC concrete. In particular, machine learning has not been applied to quantify the effect of roof-tile waste (RTW) on the CS of IC concrete. This research presents an innovative hybrid model that combines random forest and particle swarm optimization (RF-PSO) to predict the CS of IC concrete using RTW as an IC aggregate. Before model building, a comparative analysis of potential methodologies was conducted, highlighting the key characteristics, benefits, and drawbacks. RF-PSO was then chosen, achieving enhanced accuracy with a coefficient of determination (R2) of 0.961, a root mean square error (RMSE) of 5.361 MPa, and a mean absolute error (MAE) of 4.001 MPa. The RF-PSO model improved prediction accuracy by increasing R2 from 0.906 to 0.961 and reducing statistical errors by nearly 30% compared with conventional machine learning models. A Shapley Additive exPlanations (SHAP) analysis was performed to interpret the model results. The analysis identified the water-to-cement ratio and curing age as the dominant predictors, while IC water contributed a secondary, age-dependent effect. The proposed framework makes contributions: it integrates SHAP-based interpretability into a high-accuracy RF-PSO model and provides a viable tool for reducing empirical trial mixes in sustainable design workflows. Despite the limited dataset, the findings provide a reproducible baseline for future expansion and highlight the potential of combining RTW with IC to improve early and long-term strength.

Adaptive fault diagnosis in renewable integrated microgrids using hybrid machine learning approach

Assessment of free vibration frequencies of nano-scale functionally graded materials using hybrid machine learning approaches

A review of hybrid computational fluid dynamics and machine learning approaches for the combustion of alternative fuels