Research On Community Detection Of Complex Network Based On Single View And Multi View

The 21 st century is an era of explosive growth of data.It is becoming more and more important to extract useful information from these massive data.Community structure is an important attribute of complex networks.With the establishment of complex networks,community detection research has attracted much attention.Traditional community detection attempts to obtain a high evaluation function through various algorithms.Modularity is a widely used evaluation function.Because of the limitation of the resolution of modularity,the result of community division is not ideal.The improved modularity incorporates adjustable parameters to avoid the limitation of resolution.In this paper,a GASA algorithm is proposed to improve the maximum modularity.GASA algorithm is a combination of genetic algorithm and simulated annealing algorithm.It has not only the global search ability of genetic algorithm,but also the local search ability of simulated annealing algorithm.Compared with single algorithm,GASA algorithm has more obvious advantages?Taking the community of Hakka traditional villages as an example,the data of geographical environment factors of Hakka traditional villages are collected and sorted out.On this basis,the data are digitized.The environmental data of each village is expressed by 12-dimensional vector.Nodes in complex networks represent traditional villages.Whether there are edges between nodes depends on the similarity of the environment between nodes,thus building complex networks.Using GASA algorithm,the complex network constructed by traditional villages is divided into four communities.Traditional single view data is relatively simple,and single data may cause unsatisfactory results of community partitioning due to its own reasons.Multi-view data can complement each other and improve the clustering effect.Many practical problems also depend on multi-view data,but multi-view data fusion is a very complex process.In order to deal with the problems existing in multi-view clustering,we propse AMSS algorithm.Firstly,seamless integration strategy is used to integrate multi-view data,such as the attributes of topology,nodes and edges.The integration of these data improves the clustering.Class effect.Secondly,the feature attributes are mapped into high-dimensional space,and each node is given an attribute function by graph embedding.The function can recognize the non-linear features,assign a weight to each attribute,and realize the selection of non-linear features.In this way,similar nodes can be clustered together and dissimilarnodes can be pushed away.The automatic weighted multi-view model proposed by AMSS algorithm not only realizes seamless integration,but also partitions these nodes with attributes into communities.Tests on data sets show that the algorithm has certain advantages in solving multiview clustering problems.