Research On User Interest Evolution Model Based On Social Relationships

With the rapid development of mobile Internet technology,social networking is becoming increasingly popular.As one of the most popular social networking platforms in China,Sina Weibo attracts a large number of Chinese Internet users,including huge businesses,due to its timeliness and explosive propagation speed.With the increasing attention paid to the precise advertising model based on user interest we find the traditional advertising recommendation model has a low effect and easy to get bored.The huge number of users and user interests in Sina Weibo has important research value.How to accurately describe and predict user interest has important research significance.In current research,most studies are conducted from user text information or social relationships and do not describe user interests well.Since the timeliness of Weibo text information may also bring new interest to users at any time,based on this,this paper proposed an interest prediction model based on weighted WSASLA algorithm(Weighted Stochastic Approach for Link-Structure Analysis)and Ebbinghauser forgetting curve.This paper has done the following research:As for the processing of Weibo corpus texts,it is not possible to directly extract topics from LDA(Latent Drichlet Allocation)model by using the sparse Weibo text set.In view of the above situation,first of all,remove the meaningless content and webpage links in the Weibo text;secondly,using an external corpus to expand the original Weibo text set.Compared with the topic-word probability model of the direct training Weibo text and training the Chinese Wikipedia corpus,the words in the topic are more intuitive to represent the subject area in the Wikipedia corpus training results.In the study of the user interest model of social relationships,on the basis of training Weibo text theme through the LDA theme model,considered the influence of the user's interest in the user relationship on the user's interest,using extends the topic of a high-impact user to a user's interest by weighting the strength of a social relationship.In the process of analyzing the strength of the relationship between users,this paper divided the social relationship between users into strong and weak relationships according to the relationship of concern and interaction,and given the definition of the strength of the relationship and the implementation of the algorithm.In the process of user influence analysis,this paper put forward a kind of improved WSALSA algorithm based on social relations to sort the influence of user nodes in the network structure,through the research and analysis of Weibo social network.The experimental results and the evaluation results showed that the improved ISASLA algorithm proposed by the paper has a good effect on sorting results.To describe changes in user interest over time by dividing the time sequences into time slices,this paper proposed a social relationship-topic mapping model weighted by social relationship strength to consider the influence of users on the interests of other users.Compared the topics in the same time window with followed influential users,expanded the user's original interest topics with relationship weighted topics.Finally,through experiments and evaluations find that the user's topic obtained through WSALSA method have a more comprehensive user interest performance.In the current research on the evolution model of user interest,the evolution of interest is mainly analyzed through the change of user interest intensity and the change of content.Therefore,this paper used Ebbinghaus' forgetting curve law and repeated learning process to map to successive user time windows in the form of multi-stage forgetting curves.By analyzing the probability changes of user topics in time slices,we analyze the changes of user interest.Analyze changes in user interest by probabilistic changes in user topics in each time window.