Feature Relevance Analysis and Classification of Road Traffic Accident Data through Data Mining Techniques

This research work emphasizes the significance of Data Mining classification algorithms in predicting the factors which influence the road traffic accidents specific to injury severity. It precisely compares the performance of classification algorithms viz. C4.5, CR-T, ID3, CS-CRT, CS-MC4, Na ve Bayes and Random Tree, applied to modelling the injury severity that occurred during road traffic accidents. Further we applied feature selection methods to select the relevant road accident related factors and Meta classifier Arc-X4 to improve the accuracy of the classifiers. Experiment results reveal that the Random Tree based on features selected by Feature Ranking algorithm and Arc-X4 Meta classifier outperformed the individual approaches. The results have been evaluated using the accuracy measures such as Recall and Precision. In this research work we used the road accident training dataset which was obtained from the Fatality Analysis Reporting System (FARS), provided by the University of Alabama's Critical Analysis Reporting Environment (CARE) system.