Cluster analysis of crashes on intersection types in Belgium

Abstract

A great number of road traffic crashes or accidents occur at intersections worldwide. In the year 2005, about 34% of road crashes in Belgium occurred at intersections alone. Due to the complexity of road crashes, the crash data collected is usually very large and heterogeneous. Data mining techniques have become increasingly useful to analyse this type of data in order to reduce its heterogeneity and discover obscure patterns. The main objective of this study is to determine the dominant crash or accident types at various types of intersections in Belgium using cluster analysis. The purpose of cluster analysis is to group similar observations or objects; in this case crashes or accidents into homogenous groups or clusters from which meaningful insights can be obtained. Distance-based clustering techniques including K-means and hierarchical clustering and also fuzzy clustering are first reviewed and their strengths and weaknesses highlighted. These traditional distance-based clustering