Research On The Social Network Analysis-Oriented Data Mining Method

Since the emergence of the computer network, the computer has become a new kind of communication tools, and this social interaction and communication has builded a virtual space. In this virtual space, the users use Internet to build their own networks of relationships. Therefore a social network has builded in a virtual space. The study of criminal networks is considered as a specific application in social network and after September Eleven, the security sector in various countries began to attach great importance to crime data collection and associated database construction as well as a series of studies in data mining based on relational databases, which also made the study of the criminal network had been widely to expectations achieving in combating crime and maintaining public security.Social network analysis-oriented data mining method is the basic and important function in social network analysis, so is it in criminal network analysis. This paper attempts to study this problem on several aspects, and the contents and main contributions of this paper are as follows:Basic Relevance SearchingThis paper defines the concept of the basic relevance, designed and implemented based on basic relevance multi-degree relationship mining algorithm, and completed a Public Security Information Multi-Degree Relationship Mining System in Zhejiang Province Public Security Information System Platform. The system currently supports the function of mining nine kinds of basic relevances, including living together, countrymen, travelling by the same flight, and so on. The system is indeed effective to search for suspects and confirm the relationship between the suspects. In the realization, the system will not only take advantage of multi-thread technology ensuring the mining performance, but also graphically display the results of mining.Implicit Relevance SearchingThis paper defines the concept of the implicit relevance, proposed and implemented a new algorithm to find the implicit relevances in a dynamic network. The algorithm first determines a set of core features in a series of event data, and then tracks the evolution of the core features value in order to obtain the relevance network of the event sponsor, that is, the implicit relevance network. Further, through filtering the relevance whose weight value is not high, network core can be obtained. A global terrorist attacks data were used to validate the algorithm, and the result showed the algorithm is effective.Perceptual Relevance SearchingThis paper defines the concept of the perceptual relevance, proposed and implemented an implicit collaboration method for searching perceptual relevance, and tested the performance of three recommendation method, Count Method, Natural Logarithm Weighting Method, and Majority Fuzzy Preference Method, with varied time attenuation coefficient on the data set from an e-commerce site. This series of work point a direction for the public security application.