Research On Road Traffic Safety Situation Based On Fuzzy Clustering Algorithm

The frequent road accidents are serious threat to people’s lives and property in our country. Road traffic safety situation in different areas are clustered to find out similar traffic safety areas, profile the traffic safety features to make a contribution to traffic safety control and prevent road traffic accidents. This paper analyses road traffic safety situation in China from the view of classification, the characteristics and the influence factors. Some generic problems of different areas macro traffic safety is proposed, including clustering index selection and importance, number of clusters, boundaries between the class and the class. The paper is going to solve the problems from these aspects.Firstly, fuzzy set theory is introduced to solve fuzziness and intermediary of regional traffic safety situation belonging to the special class. The two fuzzy clustering algorithms are mainly analyzed:fuzzy clustering algorithm based on equivalence relation and fuzzy k-means clustering algorithm. The advantages and disadvantages of the two fuzzy clustering algorithms are summarized on analyzing road traffic safety.Secondly, the optimal clustering number and clustering center which will be regarded as initial clustering number and clustering center of fuzzy k-means clustering algorithm are determined by F statistic. The purpose is to solve sensitivity to the initialization of fuzzy k-means clustering algorithm. The road traffic accident analysis relative indexes are used as fuzzy clustering attributes and weighting theory is introduced in fuzzy similar matrix to distinguish the importance of different attributes.Finally, the paper summed up the two aspects above and put forward the new algorithm to analyze road traffic safety clustering: fuzzy k-mean clustering algorithm based on equivalence relation. It analyzed road traffic situation of31provinces municipalities and autonomous regions in China, comparing with cluster results of other fuzzy clustering algorithms. The purpose is to prove the validity and advantage of the new theory. The paper evaluates algorithm cluster results and gives the suggestions about traffic management and control.