Ranking Analysis of Highway Accident Impact Factors Based on Machine Learning Methods

Abstract

Road traffic accidents occur frequently recent years. To improve the driving safety of drivers, traffic accident analysis attracts road safety management agencies to investigate what lead to accidents and how to prevent them. In this paper, the impact factors of traffic accidents were analyzed and ranked by using field data of the year of 2012 in California from the Highway Safety Information System (HSIS) dataset. Accident severity was taken as the dependent variable; and thirteen indexes, which are related to environment, roads and drivers, were selected as impact factors respectively. Firstly, the dataset of the accident causation analysis was established after data cleaning. Secondly, support vector machine-recursive feature elimination (SVC-RFE) and random forest-recursive feature elimination algorithm (RF -RFE) were used to rank the importance of indicators. After triple cross validation, the optimal number of features of the two algorithms is obtained. The results show that the rankings obtained by the two methods are similar whether the accident severity or the degree of casualties are the dependent variables. For RF -RFE with high cross-validation scores, the similarity rate can reach 66.7%. The results of this paper can lead to traffic accident analysis methods and can guide effective prevention measures.