Modeling and Prediction of Bus Operation States for Bunching Analysis

Abstract

Bus bunching deteriorates transit service quality and passengers’ experience. The modeling and prediction of bus operation states are essential for improving the quality of transit service. Due to the nature of traffic evolution and state transition, bunching-oriented modeling based on bus operation state is more intuitive when compared with the headway-based modeling approach. This work explicitly predicted bus operation state by modeling the dynamic evolution of different states. Five different bus operation states were defined and classified by the K-means algorithm, and the dynamic state evolution was formulated as a Markov chain model. Finally, a multinomial logistic model was developed to predict the bus operation state. A case study was designed to test the performance of the proposed model based on the Global Positioning System (GPS) trajectory data collected from four bus routes in Xi’an, China. The results showed that the proposed model was able to accurately predict the bus operation states.