Modeling and Predicting Group Activity over Time in Online Social Media
HT '09: Proceedings of the Twentieth ACM Conference on Hypertext and Hypermedia, New York, NY, USA, ACM, (Juli 2009)Zusammenfassung
This paper develops a probabilistic framework that can model and predict group activity over time on online social media. Users of social media sites such as Flickr often face the enormous challenge of which group to choose, due to the presence of numerous competing groups of similar content. Determining an empirical measure of significance of a group can help tackle this problem. The proposed framework therefore determines an optimal measure per group based on past user participation and interaction as well as likely future activity in the group. The framework is tested on a Flickr dataset and the results show that this method can yield satisfactory predictions of group activity. This implies that the computed measure of significance of a group can be used by end users to choose groups with rich activity.
