Real-Time Prediction of Bus Inter-Stop Travel Time Using Deep Learning Approach

Urban bus systems are becoming increasingly important as urbanization and traffic volumes increase. Travel time is an important component of this system. Providing accurate information about the future is essential for improving user satisfaction and optimizing the use of existing facilities. With the implementation of automatic vehicle location (AVL) systems for monitoring bus locations, it is possible to access bus traffic data, which is helpful for forecasting. Analyzing AVL data from Tehran, Iran, this study compares a statistical approach to a deep learning approach for predicting inter-stop travel time. According to the results, deep learning outperforms the statistical model in travel time prediction. Additionally, the sensitivity analysis shows that arc lengths and directions are the most significant factors in travel time predictions. The developed models can predict travel times in transit applications with reasonable accuracy. Developing countries with similar public transportation systems and mobility characteristics can use the findings to improve bus services.