115
6

Distributionally Robust Prescriptive Analytics with Wasserstein Distance

Abstract

In prescriptive analytics, the decision-maker observes historical samples of (X,Y)(X, Y), where YY is the uncertain problem parameter and XX is the concurrent covariate, without knowing the joint distribution. Given an additional covariate observation xx, the goal is to choose a decision zz conditional on this observation to minimize the cost E[c(z,Y)X=x]\mathbb{E}[c(z,Y)|X=x]. This paper proposes a new distributionally robust approach under Wasserstein ambiguity sets, in which the nominal distribution of YX=xY|X=x is constructed based on the Nadaraya-Watson kernel estimator concerning the historical data. We show that the nominal distribution converges to the actual conditional distribution under the Wasserstein distance. We establish the out-of-sample guarantees and the computational tractability of the framework. Through synthetic and empirical experiments about the newsvendor problem and portfolio optimization, we demonstrate the strong performance and practical value of the proposed framework.

View on arXiv
Comments on this paper