Assortment Optimization under the Multinomial Logit Model with Random Choice Parameters

摘要: We consider assortment optimization problems under the multinomial logit model, where parameters of choice model are random. The randomness in is motivated by fact that there multiple customer segments, each with different preferences for products, and segment unknown to firm when makes a purchase. This also called mixture-of-logits model. goal choose an products offer maximizes expected revenue per customer, across all segments. establish problem NP complete even just two Motivated this complexity result, we focus on assortments consisting highest revenues, which refer as revenue-ordered assortments. identify specially structured cases optimal. When does not follow special structure, derive tight approximation guarantees extend our multi-period capacity allocation problem, prove that, restricted assortments, possesses nesting-by-fare-order property. result implies can be incorporated into existing management systems through nested protection levels. Numerical experiments show perform remarkably well, generally yielding profits within fraction percent