Research And Application Of Community Mining Algorithm Based On Node Multiple Relationships

5G application technology has greatly promoted the development of multimedia deep integration.The communication pattern between media,content and users and the mobile social network have formed a new ecological network,which makes the objects with interactive relationship show multi-relationship and group nature.For example,in the mobile communication network,while multiple users form a family group based on family relations,it forms a user group based on some products,which reflects the multi relationship of user nodes,the close connection between nodes in the group and the sparse connection between nodes in the group.Therefore,studying the community structure in the new ecological network is helpful to mine the value of the new ecological network.This paper studies the community mining algorithm with node multi relationships and its application,using the characteristics of multi node relational network,the similarity measurement index is proposed,and two community mining algorithms are designed.As the application of the algorithm,the community structure of some applications is mined on the mobile communication network.The main research contents are as follows:(1)Based on node similarity and node reachability,the path similarity-based metric LHN-IL is described.In general,similarity and path both describe the community properties between two nodes.For the locality and globality of nodes,based on the common neighbor of nodes,considering the influence of node degree at both ends and all paths of nodes,a formula for calculating the similarity of multi relational nodes is given.(2)Combined with the LHN-IL index,a community mining algorithm LL-GN(LHN-I Length-Girvan Newman)based on the improved GN algorithm is proposed to improve the quality of GN algorithm in community division The algorithm uses LHN-IL index to describe the preset low-density network model,and combines GN algorithm to divide communities.It is verified on the real data set,and the modularity Q?NMI and ARI measure indexes of the GN algorithm and Louvain algorithm are compared.The results show that the LL-GN algorithm has better community division effect.(3)By improving the LHN-IL index,a community mining algorithm LSL-GN based on node multi-relationship is proposed.In order to measure the value of the similarity index LHN-IL between two nodes without common neighbors,considering the multi relationship of nodes and the influence of increasing edge weight,a new path similarity measure index LHN-ISL is characterized.Firstly,LHN-ISL index is used to mine and reconstruct the network model with invisible similarity;Secondly,the community is divided by GN algorithm;Finally,it is analyzed and verified on the real data set.The results show that the quality of community division of LSL-GN algorithm is relatively high.(4)As the application of the algorithm,in a kind of mobile roaming network model based on "user-application" is divided into application package communities based on applications such as Ctrip Travel,Gaode Map,Didi Dache,etc.The research results provide strategic reference information for operators to formulate personalized package services.(5)Design and development of community mining visualization platform.A procedural display platform for community formation is designed based on Python and Flask frameworks.