LPA Algorithm Improvement And Strong Relationship Community Detecting On The Multi-dimensional Social Network

Community discovery is a hot issue in current complex network research and has important applications in many fields such as business,sociology,and life sciences.Many effective community discovery algorithms are proposed in the research.The tag propagation algorithm has become an important algorithm in community discovery due to its near-linear running time and ease of implementation.It is especially effective for large-scale community discovery.The traditional community discovery algorithm mainly aims at an independent network for community discovery.But in reality,due to different attributes or behaviors,the same members may belong to multiple independent social networks.For example,the students in the school are a fixed group of students.These students form a network because they belong to different classes.Another network was formed due to the different situations of participating in the community,and a network was formed due to the interaction of friends.We leave these members unchanged,and the multiple independent social networks formed by the different relationships among members are called multidimensional social networks.Community discovery based on multi-dimensional social networks(abbreviated as multi-dimensional community discovery)attempts to find strong-relational communities,that is,members in strong-relational communities all belong to the same community in multiple independent networks,and there is a high degree of homogeneity among members of strong-relational communities.In the application of advertising and news recommendation,it can make the network push service more accurate and effective.This article has carried out work in two aspects:(1)Improved the existing community discovery algorithm LPA(Label Propagation Algorithm)and formed a new node influence label propagation algorithm NFLPA(Node Influence Label Propagation Algorithm).Through experimental comparison,NFLPA has a certain improvement in the stability and accuracy of findings in the community,laying a good foundation for further research on multi-dimensional community discovery algorithms.(2)Design different multi-dimensional community discovery algorithms for different network topologies and network properties.For the case of small number of networks and large number of internal edges,a strict conditional multi-dimensional community discovery algorithm was designed.For the case of a large number of networks and a small number of internal edges,a loose-condition multi-dimensional community discovery algorithm was designed;for each dimension network In the case of unknown importance,a multi-dimensional community discovery algorithm with weighted edge loose conditions was designed.For each dimension,the importance of the network was different.A multi-dimensional community discovery algorithm with weighted edge loose conditions for dimension differences was designed and proved by experiments.A multi-dimensional community discovery algorithm can achieve effective community discovery results.