Research On Influence Communication And Evolution Tracking Technology Of Overlapping Communities In Social Networks

With the rapid development of network science and technology,as well as the wide use of mobile intelligent terminal devices,online social networks have changed people's traditional way of communication and cooperation,broken the time and space boundaries of interpersonal communication,expanded the scope of interpersonal communication,greatly met people's emotional communication needs,and even changed people's self and social cognition to a great extent.In order to analyze and utilize social network data more effectively,data mining techniques for social networks arise at the historic moment.The research of overlapping community influence propagation and evolution tracking technology in social networks is a focus of network security work,and it is also a research hotspot of social network data mining.Through the analysis of the huge user group and a large number of real-time data of social networks,it can accurately tap the potential influence of users,accurately detect the complex and diverse community structure in dynamic networks,effectively and dynamically analyze social networks information,and efficiently and accurately realize the discovery of user interest community,Accurately depict the whole life trajectory of the community,and accurately predict the content in line with the user's interest.At the same time,user influence identification,community influence maximization,overlapping community identification and evolution prediction are the key components of the research on overlapping community influence propagation and evolution tracking technology in social networks,and their performance also has a crucial impact on the overall performance of overlapping community influence propagation and evolution tracking technology in social networks.Therefore,it is of great theoretical significance and practical value to carry out research on user influence identification,community influence maximization,overlapping community identification and evolution prediction.Firstly,this dissertation deeply studies the theoretical knowledge and key technologies related to social networks,at the same time,combined with the research status in this field at home and abroad in recent years,analyzes the key problems to be solved in the current research.Then,based on the existing research results,this dissertation makes innovative research and exploration on user influence propagation,community influence maximization,overlapping community identification and evolution prediction.The specific research contents are as follows:(1)Social networks have short text,nonstandard language,contain a lot of noise data,and are easily affected by data sparsity and interest drift,which leads to low efficiency and accuracy of user interest mining and community discovery.This dissertation proposes an interest-driven overlapping network influence propagation model.First of all,considering the information transmission characteristics of overlapping network nodes and user preference characteristics,the network coupling is carried out through the "bridge" role of overlapping network users,to identify the overlapping network of the user and then based on the independent cascade model(IC),A model for maximizing the influence of overlapping networks based on user's topic preference(UI-IPM)is proposed.Then,based on UI-IPM,a two-stage mining seed node algorithm for maximizing the influence of overlapping networks is proposed.In the heuristic stage,the centrality of the node is used to initially select candidate nodes,thereby greatly improving efficiency.In the greedy stage,the greedy algorithm is optimized according to the characteristics of the sub-modules,and the seed nodes are further mined to improve the accuracy of the algorithm.Finally,the experiment verifies the effectiveness of the UI-IPM model in terms of scope of influence and time efficiency,as well as the efficiency of the IMON algorithm in mining seed nodes in overlapping networks environment.(2)To solve the problem that the existing influence models ignore the diversity of seed nodes,multi topics of community communication and the diversity of user preferences,which leads to the low efficiency and accuracy of community communication,this dissertation proposes an independent cascade model(MTL-IC)based on multi topic learning and an interest community evolution model(SPM-EE)based on similar priority mechanism.Firstly,the traditional information dissemination model designs an independent cascade model based on multi topic learning by integrating multi-topic learning factors and considering the authority and centrality of user interests.Then,according to the change of seed user's interest,seed users are updated dynamically to improve the accuracy and coverage of multi-topic information dissemination.At the same time,a community evolution model based on similar priority mechanism is designed to track the evolution process of user's interest in real time.Finally,the effectiveness of the two models in dynamic community influence maximization and multi-topic community evolution is verified by experiments.(3)In order to solve the problem that the traditional community discovery algorithm based on node neighborhood cannot effectively find the overlapping communities in social networks,first,a dynamic overlapping community discovery and evolutionary prediction model(OCDEP)is proposed.This model takes the extended label propagation algorithm(LPAE)as the semantic information model,divides the social network into user communities,and regards the user communities as the search scope of community evolution.Then,based on the findings of the LPAE community,an evolutionary prediction model based on user interest behavior,named UIBEP,is proposed.UIBEP can realize rapid and accurate community evolution by calculating the interest similarity of unlinked nodes in the community.Then,the community evolution results of each community are aggregated into a set,which is used as the community evolution result of the entire social network.Finally,the effectiveness of the two models in predicting the evolution of overlapping communities is verified by experiments.(4)Aiming at the problem of low prediction accuracy caused by existing link prediction methods that do not comprehensively consider important information such as community characteristics,text information and growth mechanisms,firstly,a user behavior preference model(UBP)based on community discovery algorithm is proposed,which improves the data quality from the source of community discovery and improves the prediction accuracy of existing link prediction algorithms.Secondly,considering the influence of important factors such as network structure attributes and user interests on link prediction,a community discovery algorithm based on improved multi-source label propagation(MSLPA)is proposed to further optimize the community structure,reduce community redundancy,and improve the prediction process.Thirdly,a friend recommendation model is designed by integrating link prediction and label propagation community discovery algorithm.Finally,a large number of experiments verify the effectiveness of the proposed model in dynamic community tracking and link prediction.