With the explosive development of big data, information data mining technology has also been developed rapidly, and complex networks have become a hot research direction in data mining. In real life, many complex systems will use network nodes for intelligent detection. When many community detection algorithms are used, many problems have arisen, so they have to face improvement. The new detection algorithm CS-Cluster proposed in this paper is derived by using the dissimilarity of node proximity. Of course, the new algorithm proposed in this article is based on the IGC-CSM algorithm. It has made certain improvements, and CS-Cluster has been implemented in the four algorithms of IGC-CSM, SA-Cluster, W-Cluster, and S-Cluster. The result of comparing the density value on the entropy value of the Political Blogs data set, the DBLP data set, the Political Blogs data set, and the entropy value of the DBLP data set is shown. Finally, it is concluded that the CS-Cluster algorithm is the best in terms of the effect and quality of clustering, and the degree of difference in the subgraph structure of clustering.