Research On Personalized Advertising Recommendation Based On Micro-Blog User Interest Model

With the rapid development of Internet technology and information communication technology, the mode of open Internet social service which is based on web2.0platform, such as micro-blog is increasingly popular. In the micro-blogging platform, everyone can express their feelings and opinions like a media. In recent years, there’re more and more researches relate to data mining based on micro-blog. In this paper, by constructing a user interest model on massive micro-blog data, we excavated the key subject words which can reveal the real interest of each user. And according to the mining results, this paper further discussed how to achieve a personalized advertising recommendation to help advertisers reduce their advertising costs and enhance the effect of advertising delivery.This article studied and explored how to use micro-blogging data to analyze users’ interests, and discussed the methods and forms of realizing personalized advertising recommendation. Compared with the existing researches of this field, this paper has put forward the following innovative points:First of all, this paper analyzed several different topic models, comparing the performance of TwitterRank model, Author-Topic model and TwitterLDA model in constructing micro-blog user interest model. According to the research contents of this paper, we finally chose the TwitterLDA model to help us identify the interest of Sina micro-blog user.Secondly, this paper applied the improved LDA algorithm to micro-blog user topic words mining. The phrases those can express the topic can be determined by analyzing the posteriori probability of topic structure. This improvement not only keeps the LDA model’s advantage of word order can be exchanged, but also improves the effectiveness of algorithm and obtains some n-gram phrases those can express the topic.Finally, this paper proposed an innovative mode of combining story concept with different advertising types in micro-blog personalized advertising recommendation and chose Sina micro-blog user as example to design an empirical research to evaluate the effectiveness of user interest model.Based on the analysis of survey results, the feasibility and effectiveness of using the topic model to dig out the real interest of ordinary micro-blog user has been verified. Besides that, the survey also includes some questions about Sina micro-blog user’s acceptance of story-based advertising, and this paper did some further discussion about it.Through this study, we find that there is a strong correlation between the micro-blog user behavior and interests, especially3main behaviors of posting, reposting and comment. The research about personalized advertising recommendation based on micro-blog user interest model is able to help us analyze micro-blog user’s interests to make the advertising delivery more accurate, so that the advertising costs can be reduced and the advertising revenue can be increased, leading to better economic and social benefits.