Personalized Modeling of Travel Behaviors and Traffic Dynamics

Abstract

Emerging mobile Internet applications have become valuable data sources for fine-grained transportation analysis, which allows the introduction of the concept of Personalization in both microscopic and macroscopic modeling of travel behaviors and traffic dynamics. Inspired by personalized recommendation systems, the personalized transportation models emphasize the importance of individual and local information. Two representative cases are presented in this study and two architectures, namely the travel behavior modeling architecture and the geoinformation modeling architecture, are proposed to address the problems of bike-sharing destination prediction and ensemble of ride-hailing demand predictors, respectively. Their performance has been verified by two case studies using the Mobike bike-sharing data and the DiDi ride-hailing demand data.