01707622

Component

https://doi.org/10.1177/0361198119851746

Modeling travelers’ mode choice behavior is an important component of travel demand studies. In an effort to account for day-to-day dynamics of travelers’ mode choice behavior, the current study develops a dynamic random effects logit model to endogenously incorporate the mode chosen for a day into the utility function of the mode chosen for the following day. A static multinomial logit model is also estimated to examine the performance of the dynamic model. Per the results, the dynamic random effects model outperforms the static model in relation to predictive power. According to the accuracy indices, the dynamic random effects model offers the predictive power of 60.0% for members of car-deficient households, whereas the static model is limited to 43.1%. Also, comparison of F1-scores indicates that the predictive power of the dynamic random effects model with respect to active travels is 47.1% whereas that of the static model is as low as 15.0%. The results indicate a significant day-to-day dynamic behavior of transit users and active travelers. This pattern is found to be true in general, but not for members of car-deficient households, who are found more likely to choose the same mode for two successive days.

The Standing Committee on Transportation Demand Forecasting (ADB40) peer-reviewed this paper (19-02414). © National Academy of Sciences: Transportation Research Board 2019.

English

pp 736-744

2019-10

Transportation Research Record: Journal of the Transportation Research Board

Web

References (36)

Cyclists; Daily; Dynamic models; Mode choice; Multinomial logits; Pedestrians; Transit riders

Pedestrians and Bicyclists; Planning and Forecasting; Public Transportation

TRIS, TRB, ATRI

Jun 4 2019 3:04PM