Balancing privacy and revenue: A differentially private dynamic pricing algorithm for ride-hailing

In this study, we propose a differentially private contextual dynamic pricing algorithm for ride-hailing platforms to address growing concerns about privacy leakage. The ride-hailing platform offers an expected price for trip requests from sequentially arriving passengers, initiating the matching process upon price acceptance. Passengers’ trip valuations are determined by a function of the trip contexts and their private preferences. While existing contextual dynamic pricing algorithms can adjust prices over time and learn optimal group pricing through individual interactions, they pose significant privacy exposure risks. Malicious third parties could potentially infer individual passenger information from price fluctuations following specific passenger engagements. To mitigate these risks, we introduce and incorporate differential privacy to design a privacy-preserving contextual dynamic pricing algorithm for ride-hailing platforms. This algorithm maximizes platform revenue, or equivalently minimizes regret relative to the online optimal policy, while ensuring bounded privacy leakage of individual passenger valuations, given knowledge of the passenger preference distribution. Theoretically, we demonstrate that the proposed algorithm satisfies ϵ-differential privacy requirements and achieves an expected regret of O˜(dTϵ), where d is the dimension of trip context and T is the total number of requests. This theoretical bound implies that as T approaches infinity, the proposed algorithm not only achieves optimality for individual passengers but also protects their privacy information almost freely.