Data-Driven Three-Wheeler Delay Prediction in Mixed Traffic conditions at Signalized Intersections Using YOLOv8 and Machine Learning

Abstract

Signalized intersections they do make things quite difficult in traffics management, particularly in developing countries like India where explicit trafficcontrolling measures often are not there. These intersections then become the hotspots for a lot of conflicts and collisions because of the lack of defined prioritymovements. Spearman's correlation analysis and ANOVA test were performed to identify the variable with a significant impact on the Delay model, which in turn influences 3W Delay. Red time, Effective width of approach, Volume per effective width, Queue length, Effective green time per cycle Time, Volume perCapacity ratio and Average Three-Wheeled Vehicle Speed at the Intersection aresignificant parameters included in this study. Using these variables as model inputs, SVR approach is employed to develop Delay models for Signalized intersections and compared with proven MLR modelling approach. Based on statistical parameters like the coefficient of determination (R2 ), the performance predictionof the best fit SVR model was evaluated.