Stochastic Approach for Price Optimization Problems with Decision-dependent Uncertainty
From MaRDI portal
Abstract: Price determination is a central research topic for revenue management in marketing fields. The important factor in pricing is to control the stochastic behavior of demand, and existing studies have tackled price optimization problems with uncertainties. However, many studies assume that uncertainties are independent of decision variables (prices), and cannot express the situation where the demand uncertainty depends on price. Although there are existing pricing studies addressing decision-dependent uncertainty, they assumed some assumptions specific to their application in order to obtain an optimal solution or an approximation solution. To handle a broader class of applications with decision-dependent uncertainty, we propose a general stochastic optimization problem by assuming more general assumptions than existing ones. While our formulation has various applications, it is hard to solve because the optimization problem is non-convex and the decision-dependent uncertainty defeats conventional theory for stochastic optimization. We focus on finding a stationary point and propose a projected stochastic gradient descent method for our problem by deriving unbiased stochastic gradient estimators. We conducted synthetic experiments and simulation experiments with real data on a retail service application. The results show that the proposed method outputs solutions with higher total revenues than baselines.
This page was built for publication: Stochastic Approach for Price Optimization Problems with Decision-dependent Uncertainty
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6510873)
