Learning to Re-Rank for Multistakeholder Recommendations

Abstract

Recommender Systems are personalized applications aiming to assist decision making by a list of tailored recommendations. Traditional recommender systems only focus on addressing the objective of the end users whose decisions ultimately determine the deal of the recommendation systems. However, the customer of the recommendations is not the only party whose needs should be represented. The needs of multiple stakeholders should be taken into account. In this paper, we are considering re-rank functions that are designed to balance end-user and online retailer needs. A variety of methodological approaches of different complexities are explored to incorporate profit information into recommenders and to balance preference and profitability. Taking the above factors into consideration, we tackle the problem balancing two sides by proposing a novel algorithm-Preference with Revenue Recommendation (PRR). Extensive experiments conducted on real world dataset indicate our method can be done with a marginal or no loss in ranking accuracy.