Access Type

Open Access Dissertation

Date of Award

January 2020

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Industrial and Manufacturing Engineering

First Advisor

Ratna Babu Chinnam

Second Advisor

Evrim Dalkiran

Abstract

Growing competitiveness and increasing availability of data is generating great interest in data-driven analytics across industries. One of the areas that has gained a lot of attention is Customer choice modeling, which aims to explain the choices individual customers make in choosing from a set of products based on their preferences. While effective customer choice modeling is essential to a wide variety of application domains, including retail, it is challenging in practice due to limitations around the quality of the data available for modeling and potentially complex choice behaviors. This dissertation presents a hybrid modeling approach that relies on both parametric and non-parametric methods to derive effective recommendations for product development and assortment planning. A generic non-parametric ranking-based choice model is first derived using random utility maximization to best model revealed product-level preferences from sales transactions and inventory records. The resulting product-level ranking-based choice model is utilized to establish customer segments and derive more actionable product attribute-based parametric models that can be employed for product assortment optimization as well as product-line extension. Then, in order to leverage from the correlatedness of customers' preferences toward similar attributes across multiple categories of products, we use cross category customer choice models to make the base predictions more accurate. The proposed modeling approach is validated using data from a leading global apparel retailer as well as synthetic experiments.

Share

COinS