Mining And Application Of Topic Diffusion Model Based On Social Network

With the rapid development of mobile devices such as smartphones,tablets and other mobile devices,social media has shown an explosive growth.Users can use their devices to publish their events or views on social media at any time and any place.The widespread use of social media makes the amount of data soaring.In order to better identity the subject of information in social media,the social media platform represents the subject of information by using hashtag(subject tag).This not only facilitates the social media on the same topic of information clustering,but also to a large extent to help users in the social network easier to find their own interest in the topic.The spread of research topics in social networks is true in many ways,such as marketing.In view of these situations,this paper studies the topic of communication in social networks.Mainly include the following aspects:First of all,according to the needs of analysis,to build a microblogging topic crawler system.To topic as the center,the breadth of the first traversal algorithm to crawl Sina microblogging topic microblogging data.Crawler system to solve the dynamic web page data analysis,as well as polite crawl data,man-machine verification and other issues,to obtain a stable crawling performance.And then crawl the data were analyzed and found that only a microblogging the number of forwarding the total number of forwarding more than 50%,indicating that the topic of microblogging data has a high degree of aggregation.Secondly,it introduces the model of information diffusion in social media-IC model and TIC model.The IC model and the TIC model assume that information is disseminated in a cascade way in a social network.According to the advantages of TIC model and the high degree of aggregation of microblogging data,this paper proposes a hierarchical topic propagation model-STIC model.The STIC model divides the topic in the social network into two layers.The first layer is the case where the information is forwarded only once,and the second layer is the case where the information is forwarded more than twice.The first layer of data to three major characteristics as input eigenvalues,the use of SVM classification algorithm learning.The second layer uses the TIC model to learn.Finally,the learning results of the two models are linearly integrated as the learning results of the whole model.Through the experimental analysis,STIC model can be better than the TIC model to predict the effect.