PROJECT TITLE :

Event Popularity Prediction Using Influential Hashtags from Social Media

ABSTRACT:

The estimation of the scope of information propagation, the making of decisions, and the prevention of emergencies all require accurate predictions of the popularity of events over social media. However, the currently available methods only concentrate on predicting the occurrences of a single attribute, such as a message, a hashtag, or an image. These methods are insufficiently comprehensive for accurately representing the spread of complex social events. In this paper, we make a prediction about the popularity of events, where an event is understood to be a collection of messages that contain multiple hashtags. We present a novel approach to predicting the popularity of events based on the influence of hashtags by mining the impact of a set of influential hashtags on the dissemination of events. Specifically, we first propose a hashtag-influence-based cascade model to select the influential hashtags over an event hashtag graph built by the pairwise hashtag similarity and the topic distribution of event-related hashtags. This model allows us to select the influential hashtags. A brand new measurement is proposed in order to determine how much of an influence a hashtag has on the content and social impacts of an event. An algorithm that is based on the correlation of hashtags is proposed as a means to optimize the seed selection in a greedy fashion. After that, we present an event-fitting boosting model as a means of forecasting the popularity of the event by incorporating the feature importance over events into the XGBOOST model. In addition, we suggest an event-structured method that can incrementally update the prediction model over social streams. This method was developed by ourselves. We have carried out a large number of experiments in order to demonstrate that the suggested strategy is both effective and efficient.