A Real-Time Bus Arrival Time Prediction System Based on Spark Framework and Machine Learning Approaches: a case study in Chiang Mai

An accurate real-time public transport prediction system occupies an important position in urban development. It includes the accurate prediction of the model and the real-time processing of the fitting data. This paper developed a bus arrival time prediction system based on the Spark framework. The process included data collection, data storage (using HDFS), data preprocessing, and modeling (ARIMAX and SVR). Moreover, we have collected data of 78 days of Chiang Mai bus real-time location and location timestamp. We used these data to construct attributes related to bus prediction. The experiment results show that the SVR model's accuracy is as high as 99.5%, which is 25% higher than that of the ARIMAX model. Therefore, the time series prediction system developed based on the Spark framework with the SVR algorithm can quickly and accurately predict bus arrival time.