Prediction of the Road Accidents Severity Using Artificial Neural Network and Comparing with Multivariate Analysis Method

Prediction of the road accidents severity based on scientific methods can be considered as an important tool in road safety management and transportation engineering. This paper presents a methodology to identify and predict the severity of accidents on highways. The methodology is based on an experimental investigation carried out on an important corridor and arterial way located in khorasan-Razavi province. The study analyses accidents occurring between 1392 and 1395 and the data concerning the accidents that were acquired from police reports. The geometric data were acquired from field study and the related office. The data, organized and stored in a specific designed Geographic Information System (GIS), were processed using a series of statistical procedures, in particular, the results took out the following two models: Model 1 was produced by Multivariate Analysis (MVA) and the Model 2 was obtained using the Artificial Neural Network (ANN) technique. Comparing the two models, it emerged that Model 2 is better than Model 1 because the total sum of the residual is lower. However, Model 1 is more efficient in estimating the black spots with a large number of accident.