Intelligent Fault Detection and Prediction in CPS Sensor Networks with Randomized Stacked Differential Decision Tree Classifier

Cyber-Physical Systems (CPS) integrate computational intelligence with physical processes and are widely used in industrial automation, intelligent transportation, IoT infrastructures, and smart grid environments. However, their reliability is frequently challenged by unexpected component failures, large-scale streaming data, and the complexities of real-time fault detection. Hardware malfunctions, software anomalies, and network disruptions can lead to significant downtime, safety risks, and financial losses. To address these challenges, this study presents an Artificial Intelligence (AI)-driven framework for accurate fault detection and proactive prediction in CPS sensor networks. The proposed Randomized Stacked Differential Decision Tree Classifier (RSDDTC) leverages differential feature analysis and multi-layer decision stacking to enhance robustness and detection accuracy while efficiently processing high-volume data streams. The model is evaluated using the BIDMC-PPG dataset, chosen for its realistic operational variations and fault-like patterns. Experimental results demonstrate that RSDDTC achieves an accuracy of 98.7%, precision of 98.2%, recall of 97.9%, and an F1-score of 98.1%, outperforming established classifiers such as Random Forest, SVM, and Gradient Boosting. Furthermore, the model reduces false detection rates by over 15%, enabling more reliable and timely corrective actions. Overall, the proposed framework offers strong scalability and real-time applicability, making it well-suited for deployment in smart grids, industrial CPS, IoT-based monitoring systems, and intelligent transportation networks, thereby contributing to the development of resilient, self-adaptive CPS solutions.