| Community discovery in complex networks is of great significance for discovering hidden rules in complex networks and predicting the behavior of complex networks.It is one of the research hotspots in recent years.Many clustering algorithms are also applied to community discovery.The density peak clustering algorithm(Clustering by Fast Search and Find of Density Peak,DPC)is a compact clustering algorithm taking density into account.Based on the DPC algorithm,this thesis proposes an automatic truncated distance parameter and cluster centers selection strategy,combining the idea of fuzzy clustering to detect the community structure of complex networks.The main work of this thesis are as follows:1.Improve the DPC algorithm based on data field and information entropy theory.Aiming at the shortcoming that DPC algorithm needs to select the cluster center artificially and different thresholds have great influence on the clustering result,the proposed method F_DPC optimizes the selection of threshold by introducing data field theory,and automatically determines the clustering centers according to the maximum information entropy reduction obtained by different segmentation methods of data sets.Meanwhile,the outlier detection condition is further optimized.2.Introduce the idea of fuzzy clustering to discover the community structure in complex network with F_DPC algorithm.Firstly,a new method is proposed to measure the distance between nodes.Then,determine the core communities with the F_DPC algorithm,and calculate the membership degree of each point with fuzzy clustering idea.Finally,complete the allocation of the remaining nodes,and set the membership threshold to distinguish the overlapping nodes.Experiments are conducted to compare the proposed algorithm with some classical algorithms in artificial data sets and real data sets,and evaluate the results through the evaluation index to verify the feasibility and effectiveness of the proposed method. |
PDF Full Text Request