The Research On Cluster Analysis For Target Group Of The Police

Along with the quick development of economy-society in China, the population is becoming more mobile, and the criminal activity is becoming groups organized. However, it is difficult for police to monitor the suspects because of the insufficient police and large population base. Thus, the implementation of strengthening the police with science and technology has become an important part of the national development strategy. Now more and more data such as relational network, call logs and trace information can be easily captured by police as the tech-information develop rapidly. Based on the cluster analysis, some tightly knit communities can be found from the big data. It is helpful for police to select the person of interest by using cluster analysis so that reduce the burden of work and find the potential criminal suspects or criminal gangs. Therefore, cluster analysis should become an important research field to solve practical problems for police.In this paper, we proposed two clustering analysis method, according to the actual needs, to solve the above problems by studying the pre-existing clustering algorithms in domestic and foreign. The first proposed method which can detect the community based on the call logs can find the crime suspects with close links to the specified crime suspects. In this method, the relationships between people are abstracted to a network after the call details being extracted, and then we can get optimal result based on the hierarchical clustering algorithm and the evaluation function of modularity. The second method can help us find the relationships between the potential suspects and the specified suspects from these trace information. In this method, the trace information such as net bars, hotel and bus station is abstracted to spatial data firstly, which is a form of one-dimensional vector. Then we extract the features from the track matrix of some people by using Latent Semantic Analysis model. Finally, clustering result can be obtained by using k-means clustering algorithm. These two methods have achieve better results in practice after cooperating with an information center of Public Security Bureau in a city.