Community Detection Research Based On Network Structure And Node Semantic Information

With the increasing scale of network data,various types of networks have become more and more complex.The community discovery methods in traditional networks face enormous challenges,and it is often difficult to solve the problem of community partitioning in high-dimensional network space.In order to more accurately discover the community structure in large-scale networks,this paper integrates related algorithms in machine learning into community discovery,and proposes an overlapping community discovery method based on network structure and node content attributes.In the process of community discovery method based on network structure,this paper first selects the DeepWalk network representation learning method to represent the nodes in the network graph as low-dimensional vectors.The low-dimensional vector representation of each node reflects to some extent the connection of the node in the network.At this point,we can treat each node as a sample,and each dimension of the vector can be regarded as a feature,and the sample data is input into the variational Gaussian mixture clustering model.Thereby the clustering result of each node in the network is obtained,and the class here is equivalent to the community in the network.In addition,the number of clusters obtained in the process can be further used as input based on content modeling.In the process of community discovery based on node content attributes,this paper uses the topic model to model the multi-dimensional attributes of the node content,so as to obtain the multicommunity attribution distribution problem of the node.The algorithm further considers the data sparsity caused by short content of the nodes in the topic modeling process.The Spike and Slab prior method is introduced in the LDA topic model to help implement variable selection and parameter estimation,then to effectively solve the sparsity and smoothness issues of community distribution on nodes.Finally,the above two methods are applied to the real network dataset through experiments.We use the DBLP literature dataset as an example to verify the above two methods.The results show that the proposed model has certain accuracy and effectiveness.Through further analysis of the excavated community structure,it better reveals the organizational characteristics within the community.