A dynamic community evolution discovery method based on attribute network representation learning

Mining the community structure in real-world networks has always been a popular topic in the complex network field. With the development of complex network research, the study of dynamic network community evolution discovery has become an important field and continues to develop. However, the current community discovery methods are mainly applied to static networks, and some methods applied to dynamic networks have the following problems: 1. The same community structure is divided into different time slices with low accuracy; 2. The time efficiency of the algorithm is low. Therefore, this paper proposes a dynamic community evolution discovery method based on attribute network representation learning. The method uses neural networks to couple the network structure and node information, obtain the representation vector of the node, then divides the community through the incremental community discovery method, finally tests the method on the artificial network and the real network. Compared with other algorithms in the artificial data set, the method in this paper has an average improvement of 4.1% in the evaluation index; in the real data set, the evaluation index has an average improvement of 3.8%, and the time efficiency has also been greatly improved.