Traffic Light Control System and Traffic Prediction Using Machine Learning with SVR and RFR

This paper examines the issue of traffic congestion in cities, highlighting its impact on higher fuel usage, loss of time, and worsening environmental pollution. The project focuses on improving traffic detection and optimizing traffic light control using machine learning methods, particularly Support Vector Regression (SVR) and Random Forest Regression (RFR)for predicting traffic and comparing their accuracy and error rates. Real-time vehicle detection and tracking are performed using Machine Learning based on video surveillance data. The methodology involves a density-based strategy where the system dynamically adjusts traffic light control, prioritizing high-traffic junctions. This is achieved using the YOLOv8 algorithm for object detection, with the system learning from real-world traffic data. The findings highlight the system's efficiency in alleviating congestion and enhancing traffic management. A comparative analysis of two regression models, Random Forest Regression (RFR) and Support Vector Regression (SVR), reveals that RFR attained an accuracy of 86.58%, while SVR slightly surpassed it with an accuracy of 87.84<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>. This indicates that both models are effective, with SVR showing a minor improvement in both accuracy and error rate. Notably, the error rate in SVR was reduced by 1.26% compared to RFR.