Abstract:

To solve the single product pricing problem of online retailers with incomplete demand information, a product pricing algorithm based on multi-armed bandit is proposed. In order to improve the effect of multi-armed bandit algorithm in the pricing problem, the algorithm takes advantage of the monotonicity of demand curve and adds consumer preference recognition. This paper analyzes the reserve price of consumers to obtain the purchase probability of consumers, models the pricing problem of online retailers into a multi-armed bandit model, presents the corresponding pricing algorithm and makes a theoretical analysis, and finally compares the pricing effect of relevant algorithms through simulation experiments. Simulation results show that the algorithm improves the revenue of online retailers.

Key words: dynamic pricing, multi-armed bandit, UCB1 algorithm

摘要：

针对在线零售商在不完全需求信息下的单产品定价问题，提出了一种基于多摇臂赌博机的产品定价算法。为了提升多摇臂赌博机算法在定价问题中的效果，该算法利用了需求曲线的单调性，并加入了消费者偏好识别。对消费者的保留价格进行分析得到消费者购买概率，将在线零售商的定价问题建模为多摇臂赌博机模型，给出了相应的定价算法并进行了理论分析，最后通过仿真实验比较了相关算法的定价效果。仿真结果表明该算法提高了在线零售商的收益。

关键词: 动态定价, 多摇臂赌博机, UCB1算法