An ANN-Based Prediction Model for Public Bus Journey Time

Abstract

Traffic congestion is a major problem in cities worldwide, especially in developing countries. It has a significant impact on the local GDP, environment, and society. Public transport is used to ease congestion but it is not efficiently implemented in developing countries. Implementing an accurate bus arrival time prediction system is a necessity to improve the standard of public transport and consumer satisfaction. In this study, we have developed a basic ANN-based prediction model for bus journey duration to estimate the bus arrival time using a case study in Johor Malaysia. The model was trained using bus fleet GPS dataset and the results shows improved accuracy compared to baseline approaches by considering factors like bus stop, time of day, month, and travel distance. Virtual bus stop is introduced for a long stretch of road and this shows promise in addressing limitations and improving performance. The simplicity of the model allows its application on any route by breaking it down into smaller segments. The final model achieves an MAE of 0.0056 and RMSE of 0.0123.