Research On Influence Maximization Algorithm Based On Dynamic User Interest Communities In A Competitive Environment

With the rapid development of Internet technology,online social networks are subverting the traditional social mode and have gradually become an important medium for people to obtain and share information,and have received more and more attention in recent years.At present,advertising,smart recommendation,online marketing,etc.can be seen everywhere on the Internet,and the inherent technologies mainly include community detection and influence maximization.However,as the scale of online social networks continues to expand,traditional algorithms can no longer meet new standards and requirements,so how to achieve algorithms with high efficiency,high precision and conformity to the real social environment has become the focus of domestic and foreign research.Based on the summary and analysis of existing community detection and influence maximization algorithms,this thesis takes into account user interest preferences,the dynamics of data and the competitiveness of information in social networks,an influence maximization algorithm based on dynamic user interest communities in a competitive environment is proposed.The proposed algorithm introduces user interest and HITS model to improve the initialization method and strategy of label propagation,and introduces a dynamic network to adjust the community structure based on the idea of incremental update.In addition,on the basis of the user communities,the user interest is introduced in the influence maximization to improve the node model,and a competitive mechanism is designed to simulate a more realistic network environment.Specifically,the main work of this thesis is as follows:(1)In view of the large amount of low-quality data in social networks,the existing label propagation algorithms are highly random,the accuracy of community detection is unstable,and the lack of timeliness of data makes the network unable to evolve,etc.,then a dynamic user interest overlapping community detection algorithm based on label propagation is proposed.The proposed algorithm first extracts user interests as labels,and selects high-quality posts and high-influence users through the HITS-based data processing model,and searches the network with the complete subgraph with high user’s centrality and topic similarity as the initial community for label propagation,and proposes the relationship strength as a basis for propagation to improve the recognition accuracy and stability of the algorithm.Then dynamically adjust the community structure based on the incremental update strategy to improve the accuracy and timeliness of the algorithm.Finally,Comparative experiments prove that compared with other algorithms,this algorithm has improved the accuracy of community detection,the accuracy of overlapping nodes and the tightness of community connections.(2)In view of the problem of the lack of authenticity in the simulation of the social environment by the propagation model and the influence maximization algorithm,influence maximization algorithms in a competitive environment are proposed.First,define the problem of influence maximization in a competitive environment.Secondly,for the propagation of competitive information,establish an independent cascading expansion model based on competition to simulate a more realistic network environment.Then calculate the influence probability between nodes based on user interest,and combined with the user interest community to calculate the comprehensive influence of nodes to restore the real social influence more comprehensively and accurately.Next,according to the different influence maximization strategy of competitive information,the problem of influence maximization in the competitive environment is divided into two categories and the corresponding algorithms IMANA and IMAPS are proposed.Finally,Comparative experiments prove that compared with other algorithms,the proposed algorithms have a greater improvement in the scope of influence.