The Research For Information Minling In Heterogeneous Soceal Networks

Social network analysis which is an important branch of data mining developed rapidly in recent years. The research of social network focus on the analysis of the relationship existed in the network to get the information that we need. The social network whose structure was single was seen as static at that time. Due to the development of the internet and computer technology, the theory of social network evolve corresponding., and heterogeneous social networks is raised, which is composed of complex and nonfigurative structure. The description of the network usually contains a variety of relationships and entities, which constitute the different structure of the network. How to process various structures and to obtain useful information of the network is a new challenge for the traditional methods of social network analysis.The research of heterogeneous social networks focused on two aspects. The first is to process the multi-relationships and integrate the traditional community mining algorithm to get eligible community structure. The other is to analyze the complex chain of relationships to obtain the information hidden in the network.This paper focuses on the research of heterogeneous social networks and proposes two algorithms based on the original algorithms.1. The algorithm of information mining based on the chain of relationships, which achieves unsupervised mining for information of network by means of analysis. The algorithm can find important information relevant to the entity or entity set. It introduced the idea of rare path that has been proved to be effectual to find hidden information in networks. To improve the efficiency and quality of the algorithm, we preprocess the given data set to get sub-graph related before analyzing, which can reduce the size of data sets and improve the efficiency of algorithm.2. We utilize the original algorithm for relation extraction and make the results of community mining more pointed. In our method, we first extract relationships of network according to tagging information of users and get the vector of correlation coefficient for the relationships. Then we introduce the idea of clustering ensemble, which regards the community division structure of every relationship as a cluster member. The idea utilize consensus function to form Co-association matrix, then combining the traditional algorithms of community mining to obtain community division structure of heterogeneous networks we want.The algorithms proposed in this article have been proved to be effective and superior by means of comparation with the original algorithm in both simulative data set and real data set.There are more and more entities and their relationships in networks due to the strengthening of informatization in every fields, that brings extensive prospects in application and new challenge in technique. The two algorithms proposed in this article can work well in application. For example, we can find canonical authors in academic network and seek marketing groups in business network according to analyzing chain of relationships. We can mark off groups of closely author in complex network of academic and obtain the circle of friends in the interpersonal network. These algorithms are more effective and high-quality than the original algorithms, though they are still immature and needed to improve further. Overall, the research of heterogeneous social networks need more attention and study, because algorithms existed of heterogeneous network can't fully meet the demand for all areas.