As the development of computer science, the commercial applications of In-Vehicle Data Recorders (IVDR) are growing rapidly. With the real-time vehicle travelling data, the implications of IVDR facilitate the rich data base of traffic accident causal analysis. Combining IVDR data analysis with driving conditions and environmental factors of different vehicles, this paper carries out the study on vehicle risk factors. On the basis of a large amount of driving and geographic location information, this study builds multiple linear regression models to analyze the relationship between vehicle risk and traffic accident, and applies empirical test to the research results. This research indicates the interact relationship between vehicle risk factors and traffic accidents, for instance, that the nonlinear relationship between travelling distance and accident probability and etc. It is expected that this study would provide useful preference for vehicle risk quantification, transport planning, and vehicle personalized pricing.