Abstract:Meetup brings people with similar interests together to do things that matter to them. For example, it provides a platform for getting people who love hiking, coding, running marathons, learning foreign languages together so that they can help, teach and learn from each other. Thanks to the development of web and mobile technologies, organizing these Meetup groups has become much more easily than before. Meetup has become an ideal tool for enriching one's social life. In this paper, we proposed a coupled linear and deep nonlinear method for Meetup services recommendation. Our method considers both historical user item interactions and group features by combining linear model with deep neural networks. In addition, we designed a pairwise training algorithm with dynamic negative sampling technique to further enhance the model performance. Experiments on two real-world datasets show that our approach outperforms the compared state-of-the-art methods by a large margin.

Coupled Linear And Deep Nonlinear Method For Meetup Service Recommendation