A system for explainable recommendations

This research aims to develop a system that generates explainable recommendations. Mobio currently allows merchants to offer items to users, but does not employ recent advances in recommender systems. This project will allow Mobio to build on the expertise of Dr. Chiang in related relational learning problems to create such a system, and provide a real-world domain for him to advance the state-of-the-art. A core problem in recommender systems is building models of user preference that are predictive and explainable. We propose to apply and evaluate algorithms developed by Dr. Chiang to build models that are optimized for accuracy, explainable to users, and recommend offers from merchants to the most appropriate users. The expected outcomes of this work include (i) enhanced user engagement through explainable recommendations, and (ii) improved effectiveness of offers by participating merchants. Addressing the interests of both users and merchants directly increases the commercial value of Mobio’s products. 

Faculty Supervisor: 
Dr. David Poole
British Columbia