Predicting the rank of trending topics

Trending topics is the most popular term list in the different web services, such as Twitter and Google. The changes in people's interest in a specific trending topic are reflected in the changes of its popularity rank (up, down, and unchanged). This paper proposes a temporal modelling framework for predicting rank change of trending topics, and delivers the real-time prediction service with only historical rank data. Historical rank data show that almost 70% of trending topics tend to disappear and reappear later. We handled those missing values, using deletion, dummy variable, mean substitution, and expectation maximization. On the other hand, it is necessary to select the optimal window size for the historical rank data. An optimal window size is selected based on the minimum length of topic disappearance in the same topic but with a different context. We examined our approach with four different machine-learning techniques using the twitter trending topics dataset, which is collected for 2 years. As an application, we implemented a trends prediction service, called TrendsForecast, applying our prediction model for Twitter trending topics in 10 different countries.