Dynamic Social Network Division Of Research Based On Node Attributes Change

In recent years, an emerging disciplines——data mining has been widespreadconcern.It was applied to the pattern recognition, statistical, database and other fields.Social networks developed quickly into a hot research topic as one applicationdirection. For the analysis of mining social network, traditional way is seen as a staticmap. It is fixed, not real.it has a lot of defects. Due to the characteristics of dynamicsocial network pays more attention to node location and the connection relations ofthe nodes varies with time, accord with nature-dynamic attributes of social network,social network analysis to predict, community identification, core node discovery, canreflect more good, so it has more research value.In this paper, with the relevant principles of complex network theory and datamining, based on social network according to the logistic curve theory, starting fromthe angle of individual initiative, dynamic social network analysis of the evolutionprocess of the relationship change attributes, and between individuals, it proposes twonovel dynamic social network partitioning algorithm which is based on node livenessand node similarity index.The dynamic social network partitioning algorithm based on node liveness forunauthorized network. Node activity description is the degree of important role ofnodes in a dynamic network in the process of evolution, the evolution process, eachnode records the neighbor node label and contact times, constructs the dynamicattribute vector for each node, and then calculate the relationship between differentnodes using the similarity function, the use of clustering number K the K-meansclustering method for small scale improvement division. The evolution process ofthe social network to consider the overall situation, the dynamic information of eachnode, in the use of static social network partitioning algorithm to partition. Dynamicsocial network partitioning algorithm based on dissimilarity between nodes inweighted networks, dissimilarity index node representation of individual intercourseor the degree of close contact, dissimilarity index expressed by the shortest pathlength, the social network is divided into different time snapshots, through the Floydalgorithm to calculate between any individual snapshot two nodes of the shortest path length size, calculate the average, and that between any two nodes dissimilarity indexsize, finally by dividing the anisotropic algorithm based on. The temporal analysismethod, based on the different time, the segmentation of the social network, to obtaina series of static images, the merge all the static diagram together to form a staticpicture of the new, then the community division in the new static maps.The algorithm is based on the analysis of the evolution process of the individual,can better reflect the characteristics of dynamic social network--dynamic.Experimental results show that the algorithm achieves good effect.