SoCaST: Exploiting Social, Categorical and Spatio-Temporal Preferences for Personalized Event Recommendations

In event-based social networks, an event recommender helps users to discover events that align with their preferences from a large number of upcoming events. In this paper, we propose a personalized event recommender called SoCaST based on the geographical, categorical, social and temporal influences of events on users to provide event recommendations. SoCaST uses an adaptive Kernel Density Estimation (KDE) to model the personalized two-dimensional geographical location. The categorical influence indicates how an event category is relevant to a user and its popularity, while the social influence is modeled as the relevance of a group to a user and her friends. Furthermore, geographical, categorical, and social influences are fused with the temporal influence which is modeled through the KDE method to generate event recommendations. Performance evaluation of SoCaST is conducted by using two large-scale Meetup.com data sets. Experimental results show that SoCaST provides better event recommendations than the state-of-the-art recommendation techniques.