一种基于多agent学习的电子市场智能定价方法

Pricing in marketplaces by multiagent learning

B2B电子市场的定价问题是一个半学习半推理的连续决策过程,每个定价agent不是直接采用多agent学习算法下的均衡策略,而是根据博弈历史进行推理决策,并不断学习对手的策略。提出了基于内省推理方法的多agent环境下agent高效在线学习方法,将基于对手模型的客观观察行为与基于换位思考推理的主观意图推测结合起来。仿真结果证实了算法在电子市场定价中的有效性。

The pricing problem in B2B electronic marketplaces is a continuous decision process,which can be seen as learning as well as inference.In stead of adopting the equilibrium policy,each pricing agent makes a decision by the pricing history of all agents.This paper proposes an efficient online learning algorithm,which integrates the observed objective behavior as well as the subjective inferential intention of the opponents,The algorithm is proven to be effective when it comes to the problem of seller＇s pricing in electronic market.