ENHANCING TRAFFIC ANOMALY DETECTION WITH A HYBRID CROW-SQUIRREL OPTIMIZATION APPROACH

Abstract

The rapid expansion of urban mobility systems has heightened the need for accurate and real-time traffic anomaly detection. Traditional approaches—including statistical models, machine learning, and deep learning—often suffer from limited generalizability, high computational cost, and reduced interpretability, particularly in dynamic environments. This study introduces a novel hybrid metaheuristic, the Crow-Squirrel Search Algorithm (CSSA), designed to enhance traffic anomaly detection through optimized feature selection and classifier tuning. CSSA combines the global exploration capability of the Crow Search Algorithm (CSA) with the local exploitation strength of the Squirrel Search Algorithm (SSA). Implemented in Python, the algorithm was tested on three benchmark datasets: UCI Traffic, METR-LA, and ShanghaiTech. It simultaneously optimizes feature subsets and Support Vector Machine (SVM) parameters. Experimental results show that CSSA achieved true positive rates of up to 95%, with a 20% reduction in false positives compared to baseline methods such as GA, PSO, CSA, and SSA. CSSA also converged faster and maintained runtime under 60 seconds per optimization run. Its robust performance across structured and unstructured datasets highlights its adaptability and efficiency. CSSA’s modular, lightweight architecture makes it well-suited for deployment in real-time intelligent transportation systems and smart city applications.