Multiobjective Deep Reinforcement Learning for Recommendation Systems

Most existing recommendation systems (RSs) are primarily concerned about the accuracy of rating prediction and only recommending popular items. However, other non-accuracy metrics such as novelty diversity should not be overlooked. Existing multi-objective (MO) RSs employed collaborative filtering combined with evolutionary algorithms to handle bi-objective optimization. Besides cold-start problem from filtering, it also vulnerable highly sparse environment, while algorithm suffers premature convergence curse dimensionality. These limitations have prompted this work propose deep reinforcement learning (DRL) approaches for MO optimization in RSs. Several works DRL available but none has addressed RS problems. In study, performances proposed that based on Deep Q-Network were investigated. The evaluated movie dataset by using three conflicting metrics, namely precision, novelty, diversity. results demonstrated superiority performance optimization, its capability precise item along achieving high against benchmark probabilistic approach (PMOEA). Although PMOEA secured higher average value lower values than approaches. surpassed maximum mean between is inevitable. addition, experiments revealed incorporation user latent features enhanced quality.