Assortment Optimization under the Decision Forest Model
Abstract
The decision forest model is a nonparametric choice model that can represent any discrete choice model, including non-rational customer behavior. This paper addresses the problem of finding the assortment that maximizes expected revenue under the decision forest model. Three different formulations are proposed, and their theoretical strengths are compared. The authors propose a methodology based on Benders decomposition to solve these problems at a large scale. The efficiency of the approach is demonstrated through synthetically generated instances. The work is joint with Yi-Chun Chen, a UCLA Anderson PhD student.
Bio
Velibor Misic is an Assistant Professor of Decisions, Operations, and Technology Management at the UCLA Anderson School of Management. He holds degrees in industrial engineering from the University of Toronto and a PhD in operations research from MIT. His research focuses on analytics, with an emphasis on customer choice, dynamic decision making under uncertainty, and optimization and machine learning intersections. He has received awards such as the INFORMS Junior Faculty Interest Group Best Paper Award and has been recognized for his excellence in research at UCLA Anderson.
