基于椭球空间表征的数据动态定价方法OA

In data trading scenarios,differentiated pricing is often pursued by data owners across heterogeneous consum-ers to maximize revenue.However,consumer valuation models are typically unobservable,and price acceptance can only be inferred from binary transaction outcomes,indicating whether a transaction was accepted or rejected.To address this challenge,a dynamic pricing method based on an ellipsoidal representation of the valuation space was proposed.Consumers'valuation functions were learned by iteratively cutting the ellipsoid,and transaction prices were subse-quently set to more closely approximate true valuations at each round.In addition,an ellipsoid cutting strategy based on parallel splits was introduced,through which valuation weights were estimated more accurately and the convergence of parameter updates was accelerated.Theoretical analysis showed that the proposed method satisfied the no-arbitrage con-straint and ensured the effectiveness of the online update process.Experiments on real-world datasets demonstrated strong performance in pricing accuracy,convergence speed,and computational efficiency.