Rasearch On Topic Based Influence Maximization Algorithm

In recent years,viral marketing based on social media networks has emerged with the rapid development of social media networks and social applications.How to choose the right initial experience users behind viral marketing to maximize the impact of product cascading in the network is a hot issue in related fields.Generally,people pay more attention to the content they are interested in.Based on this premise,choosing content that meets the interests of the people can achieve better influence transmission.In social media,there are certain message barriers between different communities due to different interests.So how to spread the news in one community to other communities and make the further diffusion of information have a greater impact is also worth paying attention to.Aiming at the above problems,this article studies a topic-based communication model and proposes a theme community-based influence maximization method.The main innovations and research work of the paper include the following:First,a TBIC model based on the topic recognition propagation model is researched and implemented.The definition of related concepts and the formal description of the propagation model based on topic recognition are given.This model uses the influence weights between users and the topic distribution of communication terms to model to represent social media networks with topic attributes.Because too many parameters of the model are prone to overfitting,the user's subject authority and interest are used to improve the model.A TAIJIC model is proposed,and a new method for processing the propagation items is proposed.Then,based on the above model,researched and implemented an influence maximization algorithm based on the theme community.First,the formal definition of the influence maximization problem based on the theme community is given.Then the main idea of the BTMG algorithm is given: select a small number of nodes from a large number of nodes as a candidate set of seed nodes,and the candidate set nodes can not only have a large impact on the propagation cascade,but also make the message spread between communities.Finally,experiments are designed and implemented.The experimental results on the Flixster and Last FM datasets verify the feasibility and correctness of the models and algorithms proposed in this paper.