Research On User Influence Of Multi-topic Social Network

Social networks play an important role in people's daily life,more and more users share content,propagation information,and express opinions on social networks,making information spread on social networks extremely active.The interconnection between users constitutes the topology structure of social networks,which is also an important channel for the propagation of information in social networks.As the main part of information,the content of social networks is indispensable in the propagation of information.Only by combining these two parts can deeply analyze the user influence of social networks.The user's interest vector can be represented by a topic distribution consisting of multiple topics.Under multi-topic conditions,the influence diffusion model of users will also be different.When analyzing the multi-topic user influence of social networks,there are two important issues: firstly,constructing an effective multi-topic influence diffusion model for global information flow,and secondly,modeling and optimizing the multi-topic influence maximization problem.In this thesis,we focus on the above two issues,and study the influence diffusion model and the influence maximization problem of multi-topic conditions.The main contributions are outlined as follows.1.Aiming at the lack of the support of the content of social networks and the limitation of topology structure in the traditional influence diffusion model,a multi-topic influence diffusion model of global information flow is proposed,and a multi-topic influence diffusion algorithm is designed.The model combines the topology structure of social networks and the topical features of content to make the model more relevant to real social networks.The concept of Topic-cluster is introduced,which breaks through the limitations of the topology structure and allows the influence to spread in the global scope.Based on the model,a multi-topic influence diffusion algorithm is designed,and the sorting list of user influence of each topic is obtained finally.In order to verify the validity of the model and the robustness of the algorithm,experiments are conducted on real social network datasets.And then,the influence of high-influence users is verified under the independent cascade model and the topic-aware independent cascade model.Experiments show that the influence diffusion model has strong topic correlation and can well simulate influence diffusion under the multi-topic conditions.In addition,the algorithm has good robustness.2.Aiming at the lack of topic correlation and poor performance of the algorithm for influence maximization problem,a multi-topic influence maximization calculation model based on maximum activation path is proposed,and a multi-topic influence maximization algorithm is designed based on the idea of the greedy algorithm.The model combines the forwarding probabilities and the similarities between users to define the activation probabilities,and uses parameters to adjust the range of influence and the accuracy of target user.Based on the topic-aware independent cascade model,the maximum activation path is introduced to quantify the spread of influence in social networks.The algorithm is optimized through data preprocessing,path filtering,and narrowing the search range.Through experiments on real data sets,a comparative evaluation of the performance of the algorithm is performed and the impact of the parameters was discussed.Experimental results show that the algorithm shows good performance and controllability.