Prediction of Market Segmentation Based on Attitudes towards Bus Trips and Risk Preference in an Urban Transit System by Bayesian Network

Abstract

Transportation policy can be more efficient in attracting a considerable number of people to choose public transit as their travel mode when decision-makers tend to develop specific policies considering different groups of people. The market segmentation method based on bus commuters’ attitude towards bus trip and their own risk preference is a significant approach to characterize various demands from bus commuters. Traditional segmentation approaches, however, rarely attempted to reveal the connection between commuters’ socioeconomics attributes and the result of segmentation due to the fact that classic market segmentation is conducted on the basis of commuters’ attitude investigation and analysis. Bayesian Network, an advanced method to make fantastic prediction, can directly predict market segmentation based on commuters’ socioeconomic attributes and risk preferences. In this way, the segmentation method can still be valid on the lack of original data of attitude and risk preference. It helps market segmentation to be more practical in demand forecasting. This paper applies Bayesian Network based on K2 and TAN structure learning algorithm respectively to predict market segmentation of attitude and risk preference on the basis of socioeconomics attributes. Traditional segmentation approach is used in this work to verify the precision of predicting segmentation results. Moreover, comparison between K2 and TAN Bayesian Network is made. The results show that the total relative error of TAN network is 29.5% while that of K2 network is 32.7%. Besides, TAN Bayesian Network takes more socioeconomic attributes into consideration than that of K2 Bayesian Network, which means the structure of TAN network coincides with common sense better. It comes to the conclusion that using Bayesian Network to predict market segmentation based on attitude towards bus trip and risk preference is capable of making the segmentation method plays a more important role in traffic demand forecasting. TAN Bayesian Network, furthermore, owns much stronger effectiveness. The proposed approach is of great help to establish potent transit systems planning and management strategies.