Analysis of the Impact of Activity on Travel Mode Choice Behavior Based on a Sequential Hybrid Method

To investigate the effects of activities on the travel mode choice, this paper proposed a sequential hybrid method aimed at capturing richer patterns in travel behavior. In the first stage, the random parameters logit model with heterogeneity in means was employed to analyze the influence of various factors and these heterogeneities. In the second stage, the XGBoost-SHAP algorithm was introduced to complement the discovery of the importance and interactions of significant factors. Results show that eleven influencing factors increase the probability of individuals choosing public transportation, among which three factors are related to activity. There are five influencing factors with random parameters, among which three factors are related to activity, including proximity of the pre-travel activity location to the nearest parking lot (≤200 m), proximity of the post-travel activity location to the nearest subway station (≤800 m), and pre-travel activity type (entertainment activity away from home). The first two influencing factors exhibit heterogeneity in the mean of random parameters. When activity-related variables interact with other attribute-related variables, they have varying effects on travel mode choices.