Research On Disjoint Community Detection Algorithm Based On Hierarchical Splitting

In nowadays,complex systems can be highly abstracted by complex networks that characterize information.The characteristics of the community structure in a complex network are its most important features so that the complex network can be mined through community detection algorithms to further study the relationship between the network structure,features,and functions.Also,the research results of community detection in complex networks have been widely used in protein complex analysis,ecological environment protection,information diffusion,and engineering technology.In recent years,a large number of community detection algorithms were proposed to obtain community division in the network.However,most of them cannot achieve good results in partition accuracy or time efficiency.This article elaborates the background of community detection and existing algorithms,and then proposes two improved algorithms based on hierarchical clustering.The main contents and contributions of this article are as follows:(1)Because of the problem that the layered split GN algorithm has poor timeliness and cannot be applied to large-scale networks,we propose a layered split algorithm JGN based on the improved Jaccard similarity coefficient.Firstly,the algorithm calculates the similarity between the nodes in the network based on the improved Jaccard similarity coefficient and guides the algorithm to cut the edge in a descending similarity manner.Secondly,the JGN algorithm continuously gives community division results in a greedy way to optimize the expansion module according to the hierarchical splitting algorithm flow.Finally,until the algorithm iteration is completed,the community division result corresponding to the maximum value of the output expansion module degree is the optimal division result.The algorithm selects and improves the Jaccard similarity coefficient to replace the marginal number in the GN algorithm during the hierarchical split iteration to improve the timeliness and stability of the algorithm.(2)To further improve the accuracy of the algorithm while ensuring good timeefficiency,we propose a hierarchical splitting algorithm SGN based on the improved Spearman correlation coefficient.The SGN algorithm refers to the Spearman correlation coefficient as a non-parametric indicator to measure the dependence of two variables and then uses the common neighbor as a basis to measure the similarity between nodes in the network to guide the operation of cutting edges during the layer splitting process.Similar to the improved method of the JGN algorithm,the SGN algorithm uses an improved Spearman correlation coefficient to replace the marginal number in the GN algorithm to improve the accuracy and timeliness of the algorithm.Finally,in order to verify the effectiveness of the algorithms proposed in this paper,the experimental part of this paper uses multiple network datasets with different characteristics to compare the performance of the proposed algorithm and the comparison algorithm on multiple indicators.The analysis of experimental results show that the community detection algorithm in this paper can efficiently detect a high-quality community structure.