Analyzing and Modeling for Mode Choice Behavior of Commuters in Metropolitan Areas

To analyze commuters’ mode choice behavior, a questionnaire was designed considering individual attributes, family attributes, and travel attributes. The nested logit (NL) model was proposed to examine commuter travel characteristics. The support vector machine (SVM) model was adopted to compare with the NL model. Traffic mode changes after policy adjustments were studied. Results showed that commuter travel time, travel costs, and transfer times are negative in the NL model coefficients, and the effect is significant. The average travel mode prediction accuracy of the NL model is 70.7%, and the SVM model is more substantial at 90.1%. The SVM model predicts the travel mode and calculates the changes after policy adjustments, respectively. The data show that the average proportion of buses, subways and trains increased by 5.68%, 0.74%, and 4.43%, and cars have decreased by 7.23% after comprehensive policy adjustment, indicating that policy adjustments can improve public transportation use.