A decision making system to automatic recognize of traffic accidents on the basis of a GIS platform

作者:

Highlights:

摘要

The prediction of traffic accidents is one of most important issues in our life. In the prediction of traffic accidents, a GIS platform to extract the important features including day, temperature, humidity, weather conditions, and month of occurred traffic accidents has been used. In this study, a decision making system (DMS) based on correlation-based feature selection and classifier algorithms including support vector machine (SVM) and artificial neural network (ANN) has been proposed to predict the traffic accidents identifying risk factors connected to the environmental (climatological) conditions, which are associated with motor vehicles accidents on the Konya–Afyonkarahisar highway with the aid of geographical information systems (GIS). Locations of the motor vehicle accidents are determined by the dynamic segmentation process in ArcGIS 9.0 from the traffic accident reports recorded by District Traffic Agency. In this DMS, firstly the number of dimension of traffic accidents dataset with five features (ay, temperature, humidity, weather conditions, and month of occurred traffic accidents) has been reduced from 5 to 1 feature by using correlation-based feature selection (CFS). In CFS method, the correlation coefficients between five features and outputs (the cases of without accident or with accident) has been calculated and chosen the feature that has highest correlation coefficient. Secondly, the traffic accident cases with one feature have been classified as without accident or with accident using SVM and ANN models. The proposed DMS has obtained the prediction accuracy of 61.79% with ANN classifier and achieved the prediction accuracy of 67.42% using SVM with RBF (radial basis function) kernel. These results have indicated that the proposed DMS could be used on prediction of real traffic accidents.

论文关键词:Geographical information systems (GIS),Traffic accident analysis,Decision making system,Correlation-based feature selection,Support vector machine,Artificial neural network

论文评审过程:Available online 6 May 2010.

论文官网地址:https://doi.org/10.1016/j.eswa.2010.04.068