conquer algorithm

Motor vehicle crashes is one of the main problems of road transportation network in Iran. Exploring the significant variables related to concentration of motor vehicle crashes is vitally important in reducing crashes on a highway corridor. This study integrates spatial analysis with a K-Mean-based Hierarchical Agglomerative Clustering method to identify the correlation between major crash types at the concentration points of crashes and explore the most spatial factors that may lead to crashes. An experiment is designed and conducted on Qazvin-Rasht highway corridor using real crash records and spatial factors related to roadway geometry and its proximity features. Results showed that clustering crash records using the proposed method is 6.7 percent better than hierarchical agglomerative clustering and approximately 10 percent better than k-mean clustering method. Moreover, analyzing the concentration points of crashes using spatial functions and discovery of patterns and rules data mining approach explored type and specification of each cluster's crashes and revealed the impact of road geometry, traffic, urban development, activities and land uses in the proximity of highway corridors on increasing the rate and severity of motor vehicle crashes.