In recent years,machine learning models have been widely introduced into spatiotemporal travel behavior modeling and prediction research due to their superior predictive performance and flexibility,but their underlying research framework and technical routes are still unclear.This article reviewed the typical literature published in related fields from 2010 to 2022 to examine the impact of the application of machine learning algorithms on the spatiotemporal travel choice behavior research paradigm,summarize the key issues to be solved in the current application and the potential influencing factors and mechanisms that affect the effectiveness of spatiotemporal travel choice behavior modeling,and foresee the directions to be focused on in future research.The effective application of machine learning algorithms to the study of spatiotemporal travel choice behavior requires not only the support of model architectures and decision mechanisms that fit the decision scenarios,but also to overcome the inherent shortcomings of all learning processes and methods,and fully consider the impact of external research conditions on the simulation and prediction performance of spatiotemporal travel choice behavior.Existing machine learning models can already fit most spatiotemporal travel choice decision scenarios,and diversified and efficient machine learning algorithms will certainly give a strong impetus to the development of spatiotemporal travel choice behavior research.Limited model interpretability remains the fundamental reason why machine learning-based spatiotemporal travel behavior models are difficult to be widely trusted.Facing the opportunities and challenges of spatiotemporal travel choice behavior research in the era of big data,it will be an important development trend to fully integrate the respective advantages of machine learning algorithms and classical decision theories and models,while improving the simulation accuracy and model interpretability of spatiotemporal travel choice behavior.
维普
万方数据
