Estimation, Prediction, and Assortment Optimization for Markov Chain Choice Model with Customization

(Joint work with Yalcin Akcay and Gerardo Berbeglia)

Abstract

We propose a framework for the Markov chain (MC) choice model with customization, including parameter estimation, customer choice prediction, and assortment optimization. In contrast to the traditional setting, which assumes that each transaction is independently drawn from a random utility model, our framework accommodates dependencies between transactions for the same customer in historical data, captured by partial-ordering preferences. To the best of our knowledge, our framework initiates the study of choice modeling with customization under MC. We present hardness and computational results for the choice prediction and assortment problem and propose novel expectation-maximization (EM) algorithms for MC parameter estimation that incorporate partial-ordering-based customization. Our EM algorithms outperform the traditional EM algorithm on synthetic datasets and the sushi dataset.

Short bio

Young-San Lin is an assistant professor at Melbourne Business School. His research interests lie in the interdisciplinary field of theoretical computer science, economics, and operations research, with a focus on market design, resource allocation, online algorithms, and revenue management. He completed his PhD in the Department of Computer Science at Purdue University.