Uncovering pedestrian midblock crash severity patterns using association rules mining
Rohit Chakraborty, Subasish Das, Md Nasim Khan
Abstract
The current study investigated the contributing factors and temporal variation in pedestrian crashes at midblock, with a particular focus on the severity levels: fatal/severe, moderate injury, and minor/no injury. It used association rules mining to uncover patterns between crash-contributing factors. By generating, evaluating, and visualising association rules for each severity level within each cluster, significant findings were discovered. Significant associations are observed between fatal crashes on weekdays and factors such as alcohol or drug impairment and nighttime. Similarly, factors including one-way roadway type, summer, and 25 MPH posted speed limit have a strong association with moderate injury crashes during weekdays. On weekends, nighttime crashes with non-motorised vehicles have the strongest association with fatal/severe injury crashes. Moreover, the generated rules for nighttime pedestrian fatal/severe injury crashes highlighted physical or drug-impaired pedestrians as the predominant attribute. The findings can enhance pedestrian safety at midblocks through targeted interventions.