Event Details

A Discussion on Semiparametric Choice Models

Discrete choice models play an important role in various applications such as route choice modelling in traffic research, product assortment and pricing in revenue management, and policy planning in health care. Most well-known choice models (e.g., multinomial logit, multinomial probit) assume that the utilities of alternatives follow a fully-known multivariate distribution. In contrast, semiparametric choice models require only partial information such as moments or marginals of this distribution. In this talk, we will first introduce semiparametric choice models rigorously and compare them to other models in the literature. Next, we will present our recent theoretical and algorithmic results for the product assortment problem under the marginal distribution model. We will conclude with a case study.