PROJECT TITLE :

Preference and Constraint Factor Model for Event Recommendation

ABSTRACT:

The recently established Event-based Social Network, also known as EBSN, is focused on bridging the gap between offline social gatherings and their online counterparts. It is becoming increasingly important for EBSNs to provide users with personalized event recommendations as the number of events that are published on these networks continues to increase. However, the vast majority of the currently available algorithms for event recommendation are unable to differentiate between the constraint factors of users' event participation behaviors and the preference factors. This reflects the cost of event participation, which prevents users from attending events in which they are interested. We differentiate preference factors from constraint factors that contribute to users' decision for event participation, and we extract the soft spatial and temporal constraints from the event venue and start time contexts, respectively, in order to take full advantage of the influences that contextual information has on users' event participation. This allows us to fully capitalize on the positive effects that contextual information has on users' event participation. Then, we propose the Preference and Constraint Factor Model (PCFM) based on the factorization machine model. This model makes use of an attentive mechanism to weight feature interactions and incorporates latent factors of users as well as contextual features in order to provide personalized preference modeling and event recommendation. In addition, learning-to-rank methods are applied in order to train PCFM to function as a ranking model appropriate for the implicit feedback nature of responses received from users. The performance of our proposed recommendation model is evaluated using extensive experiments on real-world EBSN datasets, and the results demonstrate that it outperforms state-of-the-art event recommendation methods on many metrics.