Exploring Drivers' Route Choice Behaviors in Urban Road Network by Data Mining

摘要

This study aims to investigate the rules and impact factors of drivers' route choice behaviors by mining the floating car data. The data were generated by approximately 20,000 taxies equipped with on-board GPS device in Guangzhou, China. The routes are cataloged into two types according to their origin-destination (OD) attributes: the first one is the data with the same OD pairs, that is, the origin and destination are given nodes in the same road; the second one is called the similar OD pairs, whose origin or destination are in the particular zone. The routes are further classified to investigate the route choice behavior with different choice probability. Five impact factors, travel distance, travel time, peak/off-peak period, road type, and density, are found to have significant impacts on route choice behaviors. In particular, results suggest that drivers tend to behave more stable as travel distance decreases. This study provides some perspective on the route choice behaviors analysis by data analysis, and the results provide useful information for route choice modeling such as attributes selection and expectations on the impacts of various factors.