Large-Scale Traffic Accident Data Classification Method Based on XGBoost

Abstract

Analyzing the subjective and objective of accidents causes through information from large-scale traffic accident databecomes more and more important nowadays. It is a significant reference for traffic safety prevention by researching classify large-scale traffic accident data to predict traffic accidents. In this paper, we have taken California traffic accident big data which is from US-ACCIDENT data set as our test subject, then introduces the XGBoost algorithm to classify and predict the severity of traffic accidents. The experiment results show the XGBoost algorithm looks significantly better by comparing it with several common classification algorithms. Not only in data fitting degree but also classification prediction accuracy, XGBoost works well than the other similar algorithm models. Our methods will improve the accuracy of traffic accident severity prediction to a certain extent, and provides further analysis and early warning for traffic accidents, provide an important reference for decision-making by government apparatus.