Consumption Behavior Prediction using Hierarchical Bayesian Frameworks

Purchasing products, listening to music, visiting locations in physical or virtual environments are examples of applications where users can interact with a large set of items. In this context, making predictions for both previously-consumed and new items for an individual, rather than just recommending new items, is significant in many situations. A recent work has shown that a mixture of Multinomials outperforms the widely-used matrix factorization. We further investigate this problem and propose the use of alternative mixtures based on hierarchical Bayesian frameworks to better balance individual preferences in terms of exploitation and exploration. We evaluate the alternative models accuracy in user consumption predictions using several real-world datasets, and show their efficiency for this problem.