Prediction of Re-tweeting Activities in Social Networks Based on Event Popularity and User Connectivity

摘要

This paper proposes an approach to predict the volume of future re-tweets for a given original short message (tweet). In our research we adopt a probabilistic collaborative filtering prediction model called Matchbox in order to predict the number of re-tweets based on event popularity and user connectivity. We have evaluated our approach on a real-world dataset and we furthermore compare our results to two baselines. We use the datasets crawled by the WISE 2012 Challenge (http://www. Wise2012.cs.ucy.ac.cy/challenge.html) from Sina Weibo (http://weibo. Com), which is a popular Chinese microblogging site similar to Twitter. Our experiments show that the proposed approach can effectively predict the amount of future re-tweets for a given original short message.