Leveraging machine learning to predict traffic jams: Case study of Casablanca, Morocco

Abstract

Traffic congestion in urban centers not only hampers mobility but also significantly impacts economic activities and environmental sustainability. This study leverages advanced machine learning techniques to analyze and predict traffic jams, utilizing a detailed dataset of 9847 recorded accident pixels from Waze, a popular navigation platform that aggregates real-time and historical traffic data from millions of users. The study is centered on Casablanca, Morocco and serves as a critical case study for urban traffic management. Advanced algorithms including Random Forest (RF), K-Nearest Neighbors (KNN), XGBoost, and Artificial Neural Network (ANN) were evaluated for their effectiveness in congestion prediction. The Random Forest model demonstrated exceptional performance with an accuracy of 96% and an AUC of 0.997, effectively distinguishing between congested and non-congested states. In contrast, the ANN displayed a lower performance with an accuracy of 64% and an AUC of 0.5, indicating challenges in handling complex traffic patterns. These findings underscore the potential of tailored machine learning solutions to enhance urban traffic management and planning, offering a foundation for deploying intelligent traffic systems that could significantly alleviate congestion in major urban areas.