The Research And Implementation Of Enterprise Microblogs Recommendation Based On Topic Model

The PC Internet and mobile Internet is began to integrated into people’s life, and the Internet activities of people becomes more and more frequent. As a new social platform, Microblog satisfy the people’s needs of information acquisition and daily communications.As a special kind of group of microblog, enterprise microblog can make it easier for enterprises to popularize new products, customer communication and track the industry news timely. And thus can mining more business opportunities and promote the development of enterprises. As a result of the informationoverload problem in microblogging platform, however, some users of enterprises following users too much so that many industry-related microblog are concealed by other messages, or many microblg with potential value published by other industry-related enterprise microblog user who did not followed are difficult to get. How to get industry-related microblog from a large amount of irrelevant information microblog collection, and thus make a industry situation analysis, it is important for the development of the enterprise.Traditional text mining usually based on vector space model, and the method of vector space model has its own flaws, it based text analysis only on the face value of the word, and cannot mining the potential meaning of the text, thus led to the loss of a lot of useful information. However the topic model published in recent years has the good ability in the aspect of text mining has been proven by practice, compared with the traditional method, text mining based on topic model has a better effect in finding potential topic features in text.For enterprise microblog, modeling text based on topic model is a good way to mining enterprise microblog users’industry-related interest or to distinguish microblog between different industries. It can help enterprises to get the needed industry-related information, so as to make decisions.This thesis model enterprise microblog users’topic of businesses based on topic model, using vector space model to extracted the industry feature at the same time, then establishes the industry vector of the enterprise microblog users, finally implementing the recommendation of enterprise microblog users and enterprise microblogs, this thesis studies the work embodied in the following two aspects:1. This thesis obtains sina enterprise microblog as experimental data through the BIG DATA crawler open platform, and model enterprise interest based on the LDA topic model and author-topic model ATM, experiments indicate the author-topic model ATM is superior to the LDA topic model. This is due to LDA topic model were not ideal when model the short text. Since ATM model integrated the microblog by user that eliminated the faults of text too short in a certain degree.2. After modeling the enterprise microblog user through topic model. Feature selection on all kinds of enterprise microblog. Then obtains the enterprise microblog users’ interest vector based on vector space model. Then combined the result of topic similarity calculation based on topic model and the result of industry interest vector similarity calculation based on vector space model. Finally implementing the recommendation of enterprise microblog users and enterprise microblogs based on the result of industry-related similarity calculation. The experimental results show that the proposed method has a positive effect on the recommendation of enterprise microblog users and enterprise microblogs.