Classifying the Severity Levels of Traffic Accidents Using Decision Trees

Abstract

Road accident is one of the main causes of deaths in Malaysia as well as heart disease and cerebrovascular disease. This study aims to identify the main factors that drive the occurrence of road accidents in Malaysia. Thus, preventive measures can be designed to reduce the incidence of road accidents. The relationship between the severity of road accidents and influencing factors such as vehicle movement, traffic system, marking and road geometry are also studied. The Classification and Regression Tree (CART) and Chi-square Automatic Interaction Detector (CHAID) techniques are used to identify the effects of factors in this study. The results from the decision tree show that the main factors that determine the severity of the accident are the type of vehicle, the type of violation, lighting, and severity of the driver's injuries. The performances of the two classification models are compared based on the prediction accuracy and models reliability. It is found that CHAID performs slightly better than CART and offers richer information in terms of influential factors and decision rules. The information in this study is important with the hope that road users can be vigilant and avoid being exposed to causes that allow them to be involved in accidents.