Research On Multi-objective Community Detection Problem And Algorithm In Complex Network

Various systems in life contain rich data information,which can be abstracted into network form for data mining.As one of its important research contents,community detection is of great help to the prevention and control of infectious diseases,tracking criminal organizations and other aspects,and has attracted extensive attention from experts in the fields of biology and sociology.With the network topology becoming more and more complex,the characteristics of community structure become miscellaneous,and the community detection as an optimization problem has become one of the main means in the current research.The community detection algorithm using multiple objectives can get closer to the real community,but in the research,the correlation between the objective functions is ignored,and a fixed objective function is used.At the same time,the algorithm needs to input many parameters,the local search ability is insufficient,and the convergence speed is slow,which makes it prone to a decline in accuracy when dealing with networks with complex topology structures.In order to solve the above problems,this article introduces the classification of indicators that measure the quality of community structure,studies the correlation between evaluation indicators in different network structures,and makes quantitative analysis of the correlation between evaluation indicators.By using the quantitative correlation method,after dividing the evaluation indicators,the multi-objective community detection problem is constructed according to the decision criteria for the algorithm.Considering the advantages of the gray wolf optimization algorithm,such as less parameter settings,simple implementation,complete global and local optimization strategies,fast convergence,etc.,it is suitable for dealing with the community detection problem model.This paper proposes a community detection algorithm based on multiobjective gray wolf optimization to improve the accuracy.The specific research work is as follows:(1)After introducing the classification of the current evaluation indicators of community structure quality,the research proposes the definition of the correlation between the indicators,and designs a method to quantify the correlation of the evaluation indicators according to the Spearman correlation coefficient.The verification results of some indicators on 3 real network data sets show that the proposed method can effectively measure the correlation between indicators.(2)On this basis,this paper proposes a method of constructing a multi-objective community detection problem that combines the correlation of evaluation indicators,generates a similarity matrix according to the correlation measure,and after spectral clustering,based on the highest average correlation within the cluster and the average correlation outside the cluster The lowest criterion is the evaluation index,and the result of the optimization is the multi-objective community detection problem.The experiment combined this method with 3 different multi-objective community detection algorithms on 3 real networks.The results proved that the accuracy of the 3 algorithms has been improved to different degrees.(3)A community detection algorithm based on multi-objective gray wolf optimization is proposed.In the original gray wolf optimization algorithm,in order to adapt to the research of community detection,the algorithm uses the coding based on the index of neighbor nodes,and proposes the performance verification strategy during initialization to improve the population diversity.This paper improves the leader selection mechanism in Pareto solution set,uses decision-making requirements to select leader individuals,improves the convergence speed of the algorithm,adds new individual judgment operation and external environment disturbance operation in the location update mechanism,and improves the overall optimization ability of the algorithm.The experimental results on the synthetic LFR benchmark network and the real network data set show that the detection results of this algorithm are closer to the real results than those of other algorithms.