Cluster-Based Association Rule Mining for an Intersection Accident Dataset
Mahtab Shahin, Soheila Saeidi, Syed Attique Shah, Minakshi Kaushik, Rahul Sharma, Sijo Arakkal Peious, Dirk Draheim
Abstract
Large amounts of annual costs are made for safety and compensations of accidents in urban intersections, even those with traffic lights. The main reason for accidents seems to be the convergence of different traffic flows in a particular area. The presented paper used 576 cases at intersections comprised of accident data, plus 45 fatal accident data, geometry and control status in Isfahan-Iran intersections to analyse and predict the cause of leading-to-death/injury accidents. This study used the k-modes clustering method as the main segmentation task on intersection accident data to decrease the association rule mining algorithm’s search space and remove heterogeneity of road accident data. Association rule mining helps identify the different circumstances associated with an accident in each group obtained by the k-modes algorithm. The research result shows that the extracted rules of the dataset display some valuable information that can be useful to prevent and overcome accidents.