Waf Based Community Finding And User Recommending

As the evolution of the Internet age, the web society, although virtual, is becoming an important part of the whole social communication. People participate in the virtual community, not only to consolidate his relation network of the reality word, but also to make new friends with the same interests. As a result, the social network sites have been containing abundant information about network structure and user relationship. It has becoming a more and more important studying field that how to dig out useful information from a complex social network.Based on such kind of background, this paper focus on the WAF based community finding and user recommending. It brings forward an new way to split an whole complex community to several sub communities, based on the word activity force (WAF) model. Then several experiments are designed and carried out on the user information database extracted from twitter and sina microblog, to test the algorithm. After the splitting of the whole network, a new theory to make personalized user recommend is brought forward, which aims to provide personalized recommend for user on common social network sites, such as microblog sites. Key productions of research are as follows:1. Analyze and process the user information data extracted from twitter, getting rid of useless users who may bring in serious disturb to the results, and ranking the remained users according to the HITS algorithm. Then the ranking result is discussed and analyzed in detail, as well as compared with the influence ranking given by the sites themselves.2. Bring forward the WAF based community finding algorithm. The research focus on the WAF theory’s application on user data and its change form, the processing of original user data with the WAF toolkit, the way of community finding based on the user affinity matrix, the results’ comparison and analyzing from different parameters. Besides, several different ways to deal with the isolated user node are discussed, along with their effect on the final outcome.3. Based on the results of the community finding, a modified personalized collaborative filtering algorithm is brought forward, which takes the users in a same sub community as the nearest neighbor. This algorithm takes into account the whole feature of each user, and then it helps to provide personalized recommend for user on common social network sites.Finally, the research work is summarized, and the future direction of study is discussed as well.