A Machine Learning Approach to Prediction of Passenger Injuries on Real Road Situation

According to the traffic accident automatic notification system, information on accident occurrence, location of accident, degree of injury of victims of traffic accidents is transmitted to emergency rescue agencies, which helps emergency medical treatment at the accident site and selection of appropriate medical institutions. In the paper, we propose a model for prediction of the severity of traffic accident injuries applicable to the automatic notification system of traffic accidents. We use a decision tree for the predictive model. The use data is based on traffic accident survey data and the NASS-CDS data, which collects medical records of accident victims. The predictive models developed in this paper are expected to be useful in the automatic traffic accident notification system since they have a high accurate depending on the performance analysis and supplementation of each model.