Study And Application Of Interesting Cluster Detection Algorithm Based On Link Strength

In recent years, social network entering a blowout development, is quietly changing our lifestyle and social habits. Social Networking Services is an important media for information dissemination at a great lick, people can get some important information instantly via social networking. In the face of vast amounts of information in social network, how to use effective mining methods to discover the potentially knowledge hiding in the huge data has a vital significance to the era of the information. The study found that social networks presents community structure which are similar to the real social community. Identifying the community structure accurately and efficiently is becoming hot topic in years. The traditional social network community detection algorithms has been divided into two ways, one is based on relationship data, the other is based on attribute data. The former one treats nodes as irrelevant isolated point, ignoring the relationship in information propagating process; the latter one ignoring the inherent attributes of nodes leads to lower similarity of users in the same community. Therefore, in view of the above situation, this paper discusses several classical social network community division algorithms, analyzes the idea and shortage of these algorithms. Then combined with Net Scan algorithm proposes a link strength based interesting cluster detection algorithm, and implements blog community detection on Weka which is a open source data mining platform. The main research content includes the following aspects:1.The micro-blog data are classified, put forward the notion of single user behavior and mutual user behavior used in calculating of the direct and indirect connection node link attention.2.Improve Page Rank model, proposed Indirect Link Rank model applying to the micro-blog network in order to calculating indirect connected node link attention.3.Based on the the thought of cohesive consideration node attributes and link relation,improve the shortage of Net Scan that only consider the existence of link relation,propose similar node community division based on the connection strength algorithm—LBI, use link strength in search nodes in that node in community cluster with best connection.4.Introduce the user graph interface of Weka data mining platform, focuse on the data pre-process mechanism of Weka platform, to the limitations on the handling data extend attribute filters and instances filter; Design LBI cluster using for micro-blog.