Research On Community Discovery Algorithms Based On Social Network

In the era of rapid Internet development,the network in the real world is becoming more and more complex,and the value of data behind the mining of complex networks is becoming more and more important.Community detection technology has attracted more and more attention from academia and industry.It can discover the community structure inherent in the network and has a wide range of applications in the fields of society,commerce,biology,medicine,disease prevention and control and counter-terrorism.This paper focuses on community discovery algorithms in social networks.The main work and contributions are as follows:(1)Due to the instability of label propagation algorithm in community discovery,a label propagation algorithm based on H-index(AHLPA)is proposed to solve this problem.The algorithm is based on the improved H-index to measure the influence of nodes in the network.According to the influence of nodes and extended multilayer neighbor nodes,the importance of nodes is defined,and the nodes are described in a more fine-grained way.The algorithm uses the importance of nodes to sort the update sequence of nodes,and re-optimizes the strategy of node selection tag to further reduce randomness.The experimental results show that the performance and stability of the AHLPA algorithm are greatly improved.(2)For overlapping communities,link-based algorithms have natural advantages,but there are also many problems.This paper proposes a link-based label optimization algorithm(LinkLPAm).First,the size of link-based network is generally several times larger than the original network.Therefore,the link-based coarse core is used to initialize the edge label,which not only ensures the quality of the initial solution,but also accelerates the convergence of the algorithm.Second,the idea of label propagation and optimization algorithms is combined with link-based community discovery.Finally,greedy merging of communities based on community similarity measures can avoid similar or small communities.Experimental simulation results show the effectiveness and usability of the algorithm.(3)The strategy of greedily adding nodes to local communities for general local community algorithms will fall into local optimal.This paper proposes a local community discovery algorithm(LCDGAP)based on a simple probability model.The algorithm mainly relaxes the condition that the node joins the local community,and uses the local modularity M as a measure of the tightness of the node and the community.When the node's local modularity gain ?m<0,the node is allowed to join the community with a certain probability p.At the same time,a simple probability model is proposed to balance performance and time efficiency instead of probability formula in simulated annealing algorithm.The final experimental results show that the method not only effectively reduces the parameters,but also greatly increases the Recall value of the algorithm without lowering the Precision value.(4)The user data retrieval and visualization system is designed and implemented,and the core algorithm proposed in this paper is applied in engineering practice.