Comparative Evaluation of Support Vector Machines and Random Forest Models in AI-Based Cyber-Physical Systems Control and Security Architectures

This study presents a comparative evaluation of Support Vector Machines (SVM/SVR) and Random Forest Models (RFM) for AI-based control and security architectures in Cyber-Physical Systems (CPS). This study employs a MATLAB-Simulink simulation framework to implement SVM for cyber threat classification (achieving 90% accuracy) and SVR for traffic efficiency prediction (RMSE: 0.2861), alongside Random Forest models for the same tasks. The results demonstrate that while SVM/SVR models achieved reliable performance, Random Forest models outperformed certain security detection tasks because of their improved feature importance evaluation and ensemble learning capabilities. This study analyzes model training, evaluation metrics, feature importance, and computational requirements to guide optimal model selection for real-time CPS deployment. The findings provide practical insights into AI model trade-offs in CPS applications and, support engineers and researchers in designing robust, scalable, and efficient AI-driven control and security systems.