Clustering and Classification of Road Accidents in Iran Using Data Mining Techniques

Clustering and Classification of Road Accidents in Iran Using Data Mining Techniques

Road accidents are one of the biggest public health threats in the world. In Iran, the incidence of such accidents has become more important because of the increasing number of trips and fatalities. The purpose of this study is to use data analysis techniques to extract new knowledge from data pertaining to accidents on one of the busiest roads in Iran and the factors affecting the severity of injuries sustained by vehicle drivers in these accidents. Data was collected from the traffic police database relating to accident records over a 36-month period, from January 2010 to December 2013. The authors applied clustering (Kohonen and two-step algorithms) and classification modeling (CART and logistic algorithms). The research analysis identified the eight most important causes of accidents； namely, road geometric characteristic, road direction, lane lines, lack of driver attention to the road ahead, lack of safety equipment, excessive speeding, and improper overtaking. It is hoped that this information will be used to improve overall safety by eliminating or controlling these factors.