Technology Research On Community Detection In Complex Networks Based On Intelligent Optimization

Complex network systems permeate everyone's daily life,from the Internet system to the social network.These complex networks contain the characteristics of human activities.Understanding these characteristics can guide human life,such as effective social promotion,resource allocation and so on.Community structure is an important feature of complex networks.Detecting the community structure of complex networks is an integral part of analyzing complex networks.In reality,complex networks are generally dynamic,so it is very challenging to develop algorithms for the structural relationships of communities in dynamic complex networks.It is also the mainstream direction of community detection in the future.This thesis studies the detection of dynamic complex network communities,adopts multi-objective optimization solutions,combines genetic algorithms and immune mechanisms in intelligent algorithms,and proposes a multi-objective immune dynamic detection algorithm(DYN-MDLIGA)based on double labels.DYN-MDLIGA is based on the cluster evolution framework.At the first time step,a static algorithm based on the genetic framework is used.During the initialization phase,double-layer label propagation is used to retain important node labels to ensure the stability and efficiency of the algorithm.The S-type Sigmoid function is also used to construct the adaptive cross mutation operator,which can achieve a balance between premature convergence and slow convergence speed;and adopt a multi-objective optimization strategy to avoid the resolution limitation caused by the modularity,and in the global search By adding local search optimization,experiments prove that the algorithm can achieve accurate detection and obtain a multi-level community structure.In the follow-up time steps of dynamic detection,the immune mechanism is introduced in the genetic algorithm,the immune concentration factor is set to regulate the antibody concentration,and the strategy of selecting individuals with low immune selectivity and high fitness value is established to ensure that the population is diverse and avoid premature convergence;The tag-based vaccination method achieves the goals of increasing the probability of individual selection and accelerating the rate of population convergence.Through experimental verification on artificial synthetic networks and actual real networks,the accuracy of the detection results of the DYN-MDLIGA algorithm is higher than that of the comparison algorithm while obtaining multi-level detection results.