Differential Privacy in Personalized Pricing with Nonparametric Demand Models
针对数据驱动个性化定价中的隐私问题，设计了两种不同隐私保护级别的动态定价算法，在保护客户数据的同时实现接近最优的收益最大化。
With the rapid development of artificial intelligence and big data, the application of data-driven personalized pricing has been increasingly prevalent in real practices such as finance, insurance, and retailing. However, with the public’s growing concern of the abuse of their personal data, legislation efforts are being taken to guarantee data privacy. In this work, we guarantee customers’ data privacy from the algorithm design of our dynamic personalized pricing policies. Two algorithms are developed with different levels of privacy guarantee. The first algorithm protects customers’ data in a centralized manner, meaning that the data aggregator (the pricing platform) is trusted, and the attacker is unlikely to know customers’ personal information. The second algorithm has a stronger privacy guarantee, which is mathematically proved to be able to protect customers’ data even when the data set is hacked. Besides privacy protection, both of our algorithms are effective in achieving near-optimal revenue maximization.