Research And Application Of Community Discovery Based On Improved Label Propagation Algorithm

Many complex systems in the real world can be modeled and analyzed by complex networks,and community discovery is a hot issue in complex network analysis.Community discovery can help to excavate the aggregation structure among the individuals within the complex system,analyze the association between individual and individual,grasp the law of the development of the complex system,find the hidden function of the complex system and so on,and have important research value and significance.The label propagation algorithm in community discovery algorithm is widely concerned because of its simple thought and low complexity.However,in order to solve the above problems,this paper mainly makes the following work to solve the problem of poor stability and low accuracy in the algorithm for random updating of the label.(1)This paper summarizes and analyzes the existing label propagation algorithms,and proposes an improved label propagation algorithm(Link Label Propagation Algorithm,called LLP A).The algorithm is divided into three parts.The first part is the conversion of the original network graph into the edge graph,and the edge graph keeps all the topology networks of the original network graph.The second part is to use the improved label propagation model to divide the non overlapping community on the edge map.First,the importance of the nodes is calculated and arranged in descending order,then the maximal group of the network is found by the sorting results of the nodes,as the initial community core,in the initialization stage of the label,each initial stage.The community will be given a unique label;the third part is to rerestore the edge map into a original network structure map.By traversing second parts of the community,each node in the community is restored to the two nodes of the original network,making the non overlapping community become an overlapping community discovery.Finally,the complexity and extensibility of the algorithm are analyzed,and the performance of the algorithm is improved by the experiment and comparison of artificial network and real world network.(2)This paper implements a system prototype based on community discovery of micro-blog user behavior analysis system.The system adopts the key technologies such as Spark framework,Kafka middleware,distributed storage and OLAP framework,so that the system has obvious advantages in scalability and fault tolerance.