Using Association Rule Mining to Analyze the Accident Characteristics of Intersection with Different Control Types
Yao-Tzu Hsu, Ph.D., P.E; Shun-Chi Chang

Abstract
Intersections are the most prone to accidents in the road system. Understanding the characteristics of accidents of intersections under different control types is very important for developing intersection improvement strategies. The main purpose of this research is to mine the accident influence factors and the correlation between each factor at intersections with four control types such as traffic control signal (with pedestrian signal), traffic control signal (without pedestrian signal), flashing light signal and no signal through Association Rule Mining method in data mining technology. The data source is collected from the road traffic accident data of Taichung City Police Department in Taiwan in 2017. The research results show that there are indeed differences in the characteristics of accidents at intersections of various control types. This study uses Association Rule Mining to mine the hidden information of the causes of intersection accidents, which can be used as the references for the future improvement of intersections.

Full Text: PDF