Prediksi Kemacetan Lalu Lintas Urban Menggunakan Model Pembelajaran Mesin dan Data Mobilitas Real-time

Abstract

Urban traffic congestion is a persistent challenge in rapidly growing cities, leading to increased travel times, fuel consumption, and pollutant emissions. This study aims to develop a machine-learning-based prediction model for urban traffic congestion by leveraging real-time mobility data obtained from vehicle probes and sensor networks. The proposed framework integrates supervised learning techniques including gradient boosting, random forest, and recurrent neural networks to forecast congestion levels with a lead time of 15 to 60 minutes. A dataset collected from a metropolitan region over the course of six months (including vehicle speeds, volumes, occupancy, and external factors such as weather and special events) was used for model training and validation. The results show that the best-performing model (gradient boosting) achieved an accuracy of 87% and a root mean squared error (RMSE) reduction of 23% compared to a baseline regression model. The findings suggest that real-time mobility data combined with advanced machine learning methods can significantly enhance congestion prediction performance, enabling urban traffic managers to implement proactive interventions. The study contributes to the field of intelligent transportation systems by providing a practical modelling approach and highlighting the importance of multi-source data integration. Future work should explore deployment in heterogeneous networks and test scalability across multiple cities.