The Application Of Association Rule Mining In Analysis On Property Crimes

There is a long history in the analysis of crimes. The occurred crimes may contain some information about new crimes. However, the information is not always obvious, and most of which can be presented only after the crime analysis. So the crime analysis appears. Data mining technology has become popular and has been widely applying to many fields in recent years, such as marketing, medical analysis, biological technology, bank and insurance etc. Intelligent business and web search engine are two typical examples of it. In the crime analysis field, some progresses have been made to the application of data mining, but the effects are not significant, especially on the engineering application. Based on the property crime cases of zhangzhou administrative areas, we study the application of association rule mining on them, in order to give a new method of crime association analysis, which have regional characteristics. We hope that it will be useful for the public security department of Zhangzhou city in daily works, such as criminal detection and crime prevention.Main tasks and results in this paper are listed as following:(1)Many literature are reviewed, what contain crime analysis, data mining and the combination of them at home and abroad, so as to expound their history and current situation, and then analyze the problems and difficulties in the application of data mining in crime analysis.(2)Apriori algorithm, as the core algorithm of association rules, is highlighted and analyzed in this article. For its fatal flaw, we compare and analysis many kinds of improved methods, and then we choose a kind of improved algorithm based on matrix with labels, what can be called Mtag-Apriori algorithm. Then, tests are done between the improved algorithm and the traditional Apriori algorithm, to show that the improved algorithm for mining association rules in terms of performance has obvious advantages(3)We investigate and analyze the situation of property crime in zhangzhou administrative area, and focus on the characteristics of the database, which have been established by the public security department in Zhangzhou for law enforcement. For such a large and complex database, and the demands for cross-platform, cross-regional and high reliability in law enforcement, we choose a platform named Storm, which has good performance on distributed real-time computing, to design and implement the crime association analysis system. In addition, we offer five specific rules of data preprocessing on the basis of the loose and unstructured situation in original data, and we also present a structured data mining model in line with characteristics of regional property crimes.(4) We extract numbers of property crimes from the database as a sample, to test the new method mentioned above. By adjusting the sampling strategies, and changeing the thresholds of support and thresholds of confidence, we try to enable the new method to dig up more valuable association rules. Then we explain the significance of association rules, and estimate their practical value based on the specific characteristics of the regional crimes. The results show that the new method is useful for the public security department of Zhangzhou city in criminal detection and crime prevention.