Distributed Data Distribution Mechanism In Social Network Based On Fuzzy Clustering

With the development of Internet, especially the success of social medialike Facebook, Renren, Douban etc, social network become an important partof people’s life. For social applications, one of the key problems to be solvedis how to distribute data accurately to different users and groups in highspeed. Mainstream social network services now, e.g. Facebook, Twitter,Myspace, Douban are built with concentrated services. However problemsrelated to privacy, extendibility and single point inactive are naturally withconcentrated services. Therefore, distributed social network service becomesa focal point. However, due to its own characteristics and highly dynamicuser behavior in social network, most current distributed data distributionapplications for social network do not realize functions provided byconcentrated services. In this paper, we introduce fuzzy clustering into socialnetwork analysis. Users with similar interests are clustered into the samenetwork according to fuzzy similarities. Our goal is to study how suchclustering can enhance data distribution to the largest extent. Through thecombination of unstructed topology and fuzzy presense, most events aretransmitted in the same network cluster, facilitating routing efficiency to alarge extent. At the same time, dynamically maintained user membership andinformation copies stored in multiple clusters help realize off-line file transferand guarantee its robustness in Internet-scale applications.We take10000user’s data from Douban.com and fetch fuzzy socialrelationship, simulate data transmission with a Broker-based pub/submechanism and another pub/sub system based on gossip sampling services.Experiments show that network clustering based on fuzzy clustering can improve data distribution effectively while remain robust in highly dynamicenvironment.