Prediction Method Of Microblog Public Sentiment Based On Swarm Intelligence

With the rapid development of microblog, more and more users are likely to express their opinions about some hot topics or events on microblog. The microblog includes not only the fact in the description of objective information, but also subjective information such as the author's attitude, opinion and point of view. In-depth study of these subjective information hidden in microblog can help us understand the impact of hot topic on individual sentiment state, analyze the public sentiment, and forecast the tendency of public sentiment.In order to solve the problem that most sentiment analysis methods only take the message into account, an optimization is proposed, which takes the microblog user and the user's behavior into account. For the problem of ignoring microblog users and microblog social network in microblog public sentiment prediction, which are mostly based on mathematical statistical methods, a method based on swarm intelligence is proposed. This prediction method takes microblog user's weight, sentiment fluctuation and sentiment consistency, and it includes two parts. The first part is self-evolving which represents the emotional changes of the microblog user self, and the second part is mutual-evolving which represents the emotional influences from other microblog users. The microblog user's sentiment fluctuation and sentiment consistency are based on the sentiment analysis of the history messages of the microblog user.Our experiment results indicate that the accuracy, recall and F-score can be significantly improved by the optimization at sentiment analysis. And we find a improvement of accuracy of the method proposed compared with the method based on least square.