Exploring the Role of Human in Road Crash by Analyzing Crash Variability in Similar Road Environment

Abstract: Road traffic crashes and their mitigation have always been a challenging issue. Crashes occur due to the influence and interaction between three factors: humans, vehicles, and the road environment. Due to the complexity of road crashes, it is difficult to identify the exact role these factors play in any given crash but having the information about these factors can be beneficial in developing crash mitigation plans. The human factor is often the most dominant factor responsible for crash occurrences because drivers control the vehicles. Analyzing human behavior is a complex task but assessing its interrelationship with road environment can provide road safety practitioners in-depth information about the probable causes of road crashes. The objective of this study is to explore the contribution of human behavior to the occurrence of road crashes by examining variability in crash occurrence when road environment conditions are highly similar.
In this study, a 160-kilometer mountainous semi-urban highway in Nepal was divided into 250-meter-long segments, and segments with highly similar road environments were grouped using the K-medoids clustering, an unsupervised machine learning algorithm for partitioning a dataset into clusters with similar characteristics. The clustering was carried out using the characteristics of each road segment, which consisted of road geometry, traffic condition, pavement condition, adjacent land usage, and existing road safety features. Two-year road crash data were also assigned to each road segment.
From the cluster analysis, 45 unique clusters with highly similar road environments were formed, and the variation in crash occurrence in each road segment of clusters was assessed to explore the role of human behavior for each cluster. The results showed that there exist road segments in a cluster with a high number of crashes, whereas in the same cluster, there are segments without any crashes. The crash occurrence in those segments could be attributed to human behavior because the driving conditions and existing safety measures were almost the same for all the segments. These results can be used for the prioritization and selection of interventions and can be used by road management authorities to mitigate road traffic crashes.