event

PhD Defense by Yasaman Shahi

Primary tabs

Thesis Title: Analytics for Sustainable and Retail Operations

Advisors: Dr. Beril Toktay (Scheller College of Business), Dr. Pinar Keskinocak 

 

Committee members:

Dr. He Wang

Dr. David Goldsman

Dr. Kris Ferreira (Harvard Business School)

 

Date and Time:  10:30 am-12 pm EST, Friday, July 3rd, 2020

 

Meeting URL

 

https://bluejeans.com/955110128 

 

Meeting ID

 

955 110 128

 

Abstract:

 

The vast proliferation of data has significantly transformed operational decision making in recent years. In this thesis, we present two application of analytics for non-profit and for profit organizations. 

 

Chapter 2 demonstrates an application of analytics in sustainable operations. Government regulators such as the U.S. Environmental Protection Agency (EPA) are obligated to inspect facilities regularly to ensure their compliance with environmental laws and requirements. Faced with limited budget and resources, regulators can only inspect a small fraction of facilities within certain time frame. We propose a new inspection strategy that can help environmental regulators prioritize facilities to be inspected under a limited budget. We formulate the problem as a restless multi-armed bandit model and develop an index-based inspection policy.  We also investigate how to extend the model to incorporate heterogeneous inspection costs, partial and inaccurate inspections, and the possibility of environmental disaster occurrence.  Simulations using real data from EPA show the benefits of our proposed index-based compliance monitoring strategy over other benchmark policies used in academic literature and practice in reducing the harm to the environment and public health.

 

Chapter 3 is the reflection of our collaboration with a consumer electronics retailer. The goal of this work is measuring cross-price elasticities both within product group and among competitor retailers to design a multi-product demand prediction and price optimization. In fact, we aim to show how a retailer can benefit from the historical data to make a significantly more accurate demand prediction by taking the substitution and competition effects into account.  Estimating the cross price elasticities would allow retailers to better understand the impact of price and assortment changes for products within a category on demand of other products in the category, and would ultimately help them make multi-product pricing decisions. Multi-product demand prediction models typically lack either flexibility or interpretability. We propose a neural network framework that balances these two and show how it can be used in price optimization.

 

In chapter 4, we conclude with major takeaways and scope for future work.

Status

  • Workflow Status:Published
  • Created By:Tatianna Richardson
  • Created:06/19/2020
  • Modified By:Tatianna Richardson
  • Modified:06/19/2020

Categories

Keywords