Aiming at the shortcomings such as poor scalability and low accuracy of the existing complex networks community detection algorithms,a scalable community detection algorithm of complex networks was proposed.The algorithm consisted of two phases.In the first phase,the candidate community centers of the complex networks were generated according to neighbor degree variance,similarities between nodes were evaluated through network topological structure.The similarity results were used to derive the label propagation,and the initial communities were constructed.In the second phase,the communities were fine-tuned and optimized with the help of deep reinforcement learning,both powerful sensing ability and decision ability of the deep reinforcement learning were taken advantages to improve the accuracy of communities.Experimental results show that the network communities discovered using the proposed algorithm are more accurate.
维普
万方数据