Analysis of Traffic Accidents on Highways Using Latent Class Clustering

One of the most valuable objectives of traffic accident analysis is to identify key factors of severe traffic accidents. Tracking causes or influential factors of accidents becomes possible through the analysis of large quantities of data of traffic accidents, which provides reference for lowering the number of accidents and the severity of accidents. In this paper, latent class clustering (LCC), a classification method, is applied to first classify 1,043 highway accidents, the data of which is from China’s State Administration of Work Safety, into four categories, and then to explore their respective accident attributes through analysis of seven accident variables (weather, season, time, accident category, the number of involved vehicles, etc.). This clustering of decreases its state of irregularity and homogeneity. Binary logistic regression is then employed in the whole data and the four categories to infer main causes of severe traffic accidents.