Dynamic Pricing by Multiagent Reinforcement Learning

Dynamic pricing in electronic marketplaces is a basic problem in electronic commercial. In multiagent environments, the optimal pricing policy of agent depends on the pricing policies of other agents. This makes the learning problem more problematic. This paper proposes an efficient online learning algorithm, which integrates the observed objective actions as well as the subjective inferential intention of the opponents. By establishing the decision model of other agents and predicting their proposed price in advance, agent becomes adaptive to its opponents and can make good decisions in long terms. The algorithm is proven to be effective when coming to the problem of seller´s pricing in electronic marketplaces.