원문정보
초록
영어
Thanks to the developments of recommender systems using various kinds of data, companies have adopted these systems in order to satisfy customers. However, although companies eventually want to increase their profits with recommender systems, most of recommender systems have focused on how they accurately recommend items rather than how they improve company’s sales. Accordingly, this study proposes a recommendation methodology that considers both customers’ preferences and companies’ profits. Based on the reinforcement learning model as a kind of deep learning, our methodology seeks sequentially future maximum rewards of a recommendation which is defined as profits. And, as our methodology forms a reward network, it can recommend items in real-time and to a new customer. Through experiments, we find that our methodology is better than Markov model in the perspective of expected earnings. Consequently, our methodology can be used by companies which want to earn more profit and enhance customers’ satisfaction
목차
1. Introduction
2. Related Work
2.1. Reinforcement Learning
2.2. Profit maximization on Recommender Systems
3. Methodology
3.1. Overview
3.2. Learning Phase
3.3. Recommendation Phase
4. Experimental Results
4.1. Data and Experimental setup
4.2. Results
5. Conclusion and Future Works
References