Research On Community Evolutionary Analysis Based Method Of Tensor Decomposition

With the rapid development of Internet and the advent of the Big Data era,a large number of complex networks composed of interconnected nodes and edges appear.Community is the sub-network formed by the nodes with common behavioral characteristics in the network.The dynamic changes that appear in the topological structure of a community over time are called community evolution.Excavating the community structure implied in the network and analyzing its evolution law is gradually becoming the research hotspot.At present,the research based on dynamic network mostly starts with static analysis,without considering the influence of time factor.How to effectively improve the accuracy of evolution tracking has important practical significance.Aimed at the problem that the research on dynamic weighted network is few and the influence of time factor is not considered,the paper proposes an analysis method of community evolution based on tensor model decomposition.Firstly,constructing the weighted adjacency matrix for the network snapshot of each time slice and the tensor model in combination with each time dimension;determining the number of community structure by the kernel unanimous diagnosis algorithm.Then the ATLD algorithm is introduced to decompose and reduce the dimensionality of tensor model.Secondly,theα-AOC algorithm with an adjustable parameter α is proposed to divide the community by the probability matrix of nodes belonging to the community,so that the community structure with higher accuracy can be excavated.Finally,the activity index of the community can be improved,thus analyzing the evolution law of the community structure.In order to verify the validity of the algorithm,an evaluation experiment on SocioPatterns social network datasets and Enron email datasets are conducted.The results of the community divided by the α-AOC algorithm are compared with several other mainstream community discovery algorithms,the experimental results show that the α-AOC algorithm has a more accurate result of community division.On this basis,the evolution tendency found by the community activity index agrees with the data itself,verifying the rationality of the algorithm.