Abstract

This work introduces VRNN-BPR, a novel deep learning model, which is utilized in session-based Recommender systems tackling the data sparsity problem. The proposed model combines a Recurrent Neural Network with an amortized variational inference setup (AVI) and a Bayesian Personalized Ranking in order to produce predictions on sequence-based data and generate recommendations. The model is assessed using a large real-world dataset and the results demonstrate its superiority over current state-of-the-art techniques.

Recommended Citation

Christodoulou, P., Chatzis, S. P., & Andreou, A. S. (2017). A Variational Recurrent Neural Network for Session-Based Recommendations using Bayesian Personalized Ranking. In Paspallis, N., Raspopoulos, M. Barry, M. Lang, H. Linger, & C. Schneider (Eds.), Information Systems Development: Advances in Methods, Tools and Management (ISD2017 Proceedings). Larnaca, Cyprus: University of Central Lancashire Cyprus. ISBN: 978-9963-2288-3-6. http://aisel.aisnet.org/isd2014/proceedings2017/CogScience/1.

Paper Type

Event

Share

COinS
 

A Variational Recurrent Neural Network for Session-Based Recommendations using Bayesian Personalized Ranking

This work introduces VRNN-BPR, a novel deep learning model, which is utilized in session-based Recommender systems tackling the data sparsity problem. The proposed model combines a Recurrent Neural Network with an amortized variational inference setup (AVI) and a Bayesian Personalized Ranking in order to produce predictions on sequence-based data and generate recommendations. The model is assessed using a large real-world dataset and the results demonstrate its superiority over current state-of-the-art techniques.