The Research Of Retweet Number Prediction In Social Network

With the rapid development of the Internet and the technology of Web2.0, many social network services have emerged. They are becoming the main platform of sharing, spreading and acquiring messages and gradually change the way we live. Micro-blog service is a platform of sharing, spreading and acquiring message based on users’relationship. People can post messages of up to140characters, pictures and movies through Web browsers or smart phones to share information timely. As a media platform, the message propagation is important. In micro-blog network, retweet is the main way to spread messages. Therefore, the times of retweet (i.e., retweet number) can be as an important indicator of the messages’influence. Studying the retweet behavior and predicting the scale of retweeting has practical significance in products marketing and hot extraction. What’s more, it contributes to controlling the spread of illegal information like rumors. In this paper, we study retweet behavior as follow:1.whether a tweet can be retweeted.2.predicting the retweet number of a tweet.When we study whether a tweet can be retweeted, we do it from the sentiment of the tweet and the role of the tweet’ creator. In the sentiment of the tweet, this paper constructs a text sentiment analysis engine to analyze the impact of retweet behaviors. In the role of tweet’creator, we verified that different roles have different retweet behaviors.About the scale of retweet behavior, we propose two models to predict the retweet number. One is a two-phase model for retweet number prediction. It combines the classification models and regression models to avoid the imbalance of the data, and get a good prediction. Through studying the process of the generation of retweet number, we build a retweet number prediction model based on followers’ retweet intention and influence. This model learns the process of the generation of retweet number, and has better prediction than the two-phase model.