Research On Community Detection Technology Based On Spectral Clustering

Many of the systems in the real world can be abstracted into complex networks,and community-based features are also prevalent in complex networks.Community structure provides a feasible entry point for complex network research and provides an important basis for other studies of complex networks.Therefore,community detection is one of the main hotspots in complex network research.Traditional community detection studies focus on static networks,but the network structure is changing over time in real-world scenarios.On the other hand,most of the dynamic network community detection researches are mainly on unweighted network at present,which may cause distortion of network partition.Considering the situation,the research of community detection in dynamic weighted network is very important.In this thesis,we choose the dynamic weighting network as the research object.In order to solve the problem that the process of constructing the similarity matrix is too complex in the traditional spectral clustering algorithm,we present the evolutionary spectral clustering algorithm by introducing the capocci algorithm and the random walk theory.Finally,we give the validity verification of the algorithm through the experiment of the actual data set.The main research contents include:1.We summarize the relevant concept of dynamic community detection,elaborate the basic idea of spectral clustering algorithm and graph segmentation theory,and then investigate the Yun Chi evolutionary spectral clustering algorithm and analyze the shortcomings of the algorithm in constructing the similarity matrix of weighted networks.2.In order to solve the problem of Yun Chi algorithm,we analyze the relationship among the standard cut function,the Markov chain and the transition probability matrix and use these relations to optimize the objective function of Yun Chi algorithm.As a result,it can directly deal with the weighted adjacency matrix and simplifies the process of the algorithm.3.We propose a spectral clustering algorithm to detect the community in the dynamic weighted network.The experimental results based on the cooperative network data set published by the Institute of Automation of the Chinese Academy of Sciences indicate the effectiveness of the algorithm proposed.