A Comprehensive Review of Traffic Congestion Prediction Models: Machine Learning and Statistical Approaches

Abstract

Traffic congestion prediction has become a critical component of intelligent transportation systems, enabling more efficient traffic management and urban planning. This study provides an in-depth review and analysis of the most recent methodologies applied to traffic congestion prediction, including machine learning, statistical, and hybrid models. We evaluate key studies that demonstrate how machine learning models excel at handling non-linear traffic patterns and real-time data processing, while statistical models, remain useful in stable and predictable environments. Hybrid models, which combine the strengths of both approaches, show promise in improving prediction accuracy and managing uncertainties. Furthermore, we discuss the challenges of scalability, generalizability, and the integration of external factors such as weather and accidents. Based on these findings, we propose several future directions for traffic congestion prediction research. The review highlights the necessity of further refining these methodologies to develop more adaptable and efficient predictive systems for modern urban environments.