Retweet Prediction Of Media Microblog

With the rapid development of WEB2.0 and Mobile Internet, microblogging platform has become a major source of network hot events and sensitive public opinion. How to effectively monitor and aright guide the public sentiment on the micro-blogging platform has become an issue that should be coped urgently with. Retweeting behaviors based on user relationship is the primary mode of microblog’s transmission, the number of retweeting times can measure the spread effect of micro-blog. If retweeting number can be predicted timely, the microblogs that could have a significant impact on public opinion will be founded before them break out.The microblogs posted by certified media users called media microblog. Media microblogs have high credibility and great influence, the retweeting numbers of them can reflect public concerns towards described social phenomenon. For media microblog, the direct retweeting number usually consist of the vast majority of the total number, that means the indirect retweeting behaviors can be ignored when predicting the retweeting number of media microblog. In most cases, the retweeting number of a micro-blog message grow rapidly at early phase, then will be stable after a period of time.This paper presents a method based on curve fitting and time-series model to predict the retweeting number of media microblog, the best fitting point and function are derived from the experiment. In order to improve the predicting precision, empirical correction models are built by utilizing the grouped prediction data of media microblog using polynomial regression method. The regression result of grouped data proved that the posting time has a great impact on the retweeting trend. Correction effect of different correction models and the reliability of whole predicting scheme are tested by prediction experiment on test set.In order to improve work efficiency of the public opinion monitoring on SINA microblogging platform, a surveillance software of media microblog is developed. This software can crawl the microblog data on web pages automatically and the information of microblog will be demonstrated in the form of list and graph. Data of microblog can be arranged easily by using the function of tagging, filter and sequencing. After importing the model, the retweeting number of microblog can be predicted and displayed.