Traffic congestion at urban intersections is worsening, raising demands for better traffic management and urban planning. We use machine learning models to predict vehicle stop times and distances at intersections to characterize congestion. Analyzing traffic data from Atlanta, Boston, Chicago, and Philadelphia, we select features like location, driving direction, and time, and develop five models: Linear Regression, K-Nearest Neighbors, Decision Tree, Random Forest, and Gradient Boosting Tree. These models predict the 20th, 50th, and 80th percentiles of vehicle stop times and the distance from the intersection where a vehicle first stops. Results show that Random Forest and Gradient Boosting Tree models achieve the best accuracy and generalization, effectively capturing complex traffic patterns. This research offers a valuable tool for intersection traffic management.
Data-driven prediction of intersection traffic congestion
International Conference on Smart Transportation and City Engineering (STCE 2024) ; 2024 ; Chongqing, China
Proc. SPIE ; 13575
28.04.2025
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Traffic intersection congestion prediction method based on machine learning
Europäisches Patentamt | 2020
|Traffic intersection vehicle congestion detection method
Europäisches Patentamt | 2024
|Intersection traffic control method for preventing congestion
Europäisches Patentamt | 2016
|Europäisches Patentamt | 2025
Europäisches Patentamt | 2025
|