Robust Assortment Optimization in Revenue Management Under the Multinomial Logit Choice Model

摘要: We study robust formulations of assortment optimization problems under the multinomial logit choice model. The novel aspect our is that true parameters model are assumed to be unknown, and we represent set likely parameter values by a compact uncertainty set. objective find an maximizes worst-case expected revenue over all in consider both static dynamic settings. setting ignores inventory consideration, whereas setting, there limited initial must allocated time. give complete characterization optimal policy settings, show it can computed efficiently, derive operational insights. also propose family sets enables decision maker control trade-off between increasing average protecting against scenario. Numerical experiments approach, combined with proposed sets, especially beneficial when significant values. When compared other methods, approach yields 10% improvement performance, but maintain comparable if performance measure interest.