An Analysis of Seasonal Effects on Transport Mode Choices based on GPS Trajectories and a Multinomial Logistic Regression Model

Abstract

To improve the planning of transportation modes, it is necessary to understand how passenger choose the transportation mode depending on factors such as temperature, precipitation, trip distance, trip duration. In this article, we build up on previous work to analyze the effect of seasonality in the choice of transportation modes. To that end, we use GPS trajectories in Beijing from the Geolife Trajectory data set, and analyze them with respect to the four seasons: Summer, Au-tumn, Winter, and Spring. Our analysis reveals how seasonality influences the choice of transportation mode, e.g. people tend to bike less in Summer as temperature increases, whereas rush hour traffic influences the car share in Autumn. We build a multinomial logit model, one per season, to predict the transportation mode based on the following factors: temperature, air quality, traffic during rush hour, time of day, trip distance and trip duration. The results show how changes in the factors effect the choice of transportation mode. For instance, long trips in Autumn decrease the probability of walking by 54% and increase the probability of car, bus and train use by 37.7%, 11.1% and 13.5%, respectively. We conclude that rush hours only influence the choice of transport mode during Autumn and Spring. Also, we have observed some similarities in the choice of transport modes throughout different seasons, e.g. the trip distance is a relevant factor for the choice of transport mode in all four seasons.