Prediction and Investigation of Road Traffic Accident Severity Factors using Support Vector Machine Algorithm

Accidents as a threat factor to the transport system have widespread political, social, and economic dimensions, which is increasing in developing countries. Iran, as a developing country, has not escaped this danger, but in recent years it has taken preventive measures and crash statistics has been decreasing. Crash injury severity is one of the most important criterions of measuring the costs of accidents. Different methods have been used for predict modeling. Support Vector Machine is a relatively new modeling technique which was proposed to solve the classification and regression problems. That shows accurate and acceptable performance. In the present study, it has been tried to model the severity of accidents with a combination clustering and classification approach, with the help of neural network algorithms, simple parsing, SVM, KNN, and C4.5 algorithms and by comparing the algorithms used assess the ability of each of the algorithms in the prediction of the severity of accidents. The purpose of this study is to evaluate the performance of support vector machine algorithm to predict the severity of road accidents and identify the factors that affect the severity of accidents.