In order to reduce the influence of the selection of parameters in community detection algorithm on the results of community partitioning and detect adaptively

a community detection algorithm called KDED based on kernel density estimation for density peak clustering is proposed. Firstly

a distance measure based on trust is defined

and the user relationship in the social network is quantified as a distance matrix. Then the kernel density is estimated by the distance matrix to calculate the influence of each node on the network. The thermal diffusion model improves the computational flow so that it adapts to different sizes of data sets to improve computational accuracy. The clustering center is determined by density peak clustering principle and community property

and the node is allocated to the corresponding community which can obtain the hierarchical structure within the community and the natural structure among the community. The simulation results show that it can be observed by the visualization software that the community division result obtained by the KDED algorithm has a clear natural structure and internal hierarchical structure. The KDED algorithm has the best stability with the increase in the size of the community and the difficulty of detection compared with the classical algorithm. And it gets a community partition closer to the real partition in the real data set and the LFR benchmark network

which verifies the feasibility and effectiveness of the algorithm.