Rawls College of Business Faculty Team Wins Defense Data Grand Prix

This challenge from the U.S. Department of Defense encourages academic teams to solve government problems with data science.

Ashlyn Grotegut  |  February 27, 2023

Sheikh-Zadeh (who led this project) is an assistant professor and Brown an assistant professor of practice in the Area of Information Systems and Quantitative Sciences (ISQS). Song is the associate dean of graduate programs and research and Jerry S. Rawls Professor of Management Information Systems.

Together, they make a team that represented Texas Tech University in the Defense Data Grand Prix: a challenge from the U.S. Department of Defense (DoD) to encourage academic teams to use data science and tackle compelling, real-world problems faced by the government.

This is a three-heat, 18-month competition established by the Acquisition Innovation Research Center (AIRC), which is a partnership of 22 U.S. universities. The semester-long competitions conclude with $100,000 in prizes: $40,000 first place, $30,000 second place, $20,000 third place, and $10,000 fourth place.

Heat 1 was completed in fall 2021, and Heat 2 in spring 2022. In Heat 1, teams worked with Defense Logistics Agency (DLA) analysts to recommend ways to improve access to applicable data. In Heat 2, teams focused on implementing approaches to making data accessible for analyses. The winners of those heats had their submissions published.

Heat 3 was the latest competition, in which teams applied advanced data visualization techniques on findings from the defense acquisition. They had to investigate problems such as aviation supplier predictions, manufacturing stores and materiel shortages, balancing freedom of information with operations security, and industrial capability program material identification.

The challenge began September 2022 with an orientation seminar, and all submissions were due December 15. A total of 15 teams, 24 professors, and 13 DoD representatives participated – and the Rawls College trio never dreamed they would rise to the top of the competition.

“As a scholar, we are very familiar with rejection,” Song said. “When I formulated our team, it was just building relationships and nothing more than that.”

Brown and Sheikh-Zadeh both have a background in industrial engineering. Song specializes in economics and information technology.

They first met about Heat 3 after encouragement from Rawls College Dean Margaret L. Williams. She learned about the opportunity from the Rawls Advisory Council.

“We decided that we would try it and see what happened,” Song said.

A Process From the Start

Before they could compete, each member of the team had to complete non-disclosure agreement training because of the sensitive data involved. Once cleared, they brainstormed their roles and began their teamwork.

“We said, ‘Let's go take a handful of stabs at this and see who makes progress,'” Brown said. “It was some collaboration, and some division of labor.”

The problem they began to investigate was on-time delivery prediction.

“It was actually just like predicting a rare event, like predicting credit card fraud in a dataset where only a very small fraction of transactions are fraudulent,” Sheikh-Zadeh said. “It was very challenging.”

But they pushed through the challenge in search of a solution – a task his desktop computer could barely handle.

After three weeks of feeling stumped, that relentless work paid off.

“Ali had a breakthrough, and he said, ‘I think we can win with this metric,'” Brown said.

Sheikh-Zadeh had researched spare part data before and was aware of how these kinds of supply chain work.

“If something happens, it has a trickling impact,” he said. “It has an impact on upper-level suppliers, it has impacts on lower-level suppliers, and this can also spread through the network. So, I thought about feature engineering, which means creating new features related to our prediction. In this case, it was a key moment.”

As detailed in their written submission, Sheikh-Zadeh, Brown and Song analyzed data from the DLA's order history and developed a predictive model to identify orders that would be delivered late. Their proposed repeatable framework for prediction could arm the DLA with a powerful tool to accurately identify late orders and more efficiently monitor the supply chain it oversees.

The trio went on to explain they could positively impact DLA's mission by accurately identifying late orders with minimal false positives. They believe their approach could be broadly implemented and suit the needs of DLA with its straight-forward design and computationally inexpensive methods.

They believe their approach should be considered for adoption and implementation so the DLA and DoD can garner further flexibility with taxpayer funds.

“We felt like we had a really good product,” Brown said. “We knew we were up against some intelligent people who were going to do sophisticated work. But we knew the practical implications of what we had were strong.”

Thankfully, the judging panel agreed. AIRC staff members and representatives from government sponsor agencies all gave the trio a high score, with the final award selection made by DoD officials.

“They really appreciated it,” Sheikh-Zadeh said. “They were very impressed with the amount of effort we put in and they found it fascinating to apply what we did.”

The trio was congratulated during a virtual awards ceremony on Jan. 10, but Williams could not contain her excitement from behind a screen. She, too, took to the hallways of Rawls College to celebrate with the teammates at their offices.

“I am proud of this team of faculty who took on this challenge at the last minute and successfully competed against teams from top universities across the nation,” Williams said. “Their willingness to go the extra mile is an example of the culture we promote in the Rawls College of Business and their commitment to the education of future data science students. Their first-place finish opens greater opportunities for relationships among the Rawls College of Business and defense agencies.”

Payoff in the Classroom

While it may take a while for the trio's work to be published or implemented, the Defense Data Grand Prix has already made an impact on the Rawls College.

“It was invaluable because, we as a faculty learned as well,” Sheikh-Zadeh said. “So, through this process, we really learned how we could be better professors when we teach data science and machine learning and bring that value to students.”

One lesson Brown intends to pass along is “data is messy."

As a master's in data science program faculty member, he teaches the term data scientist: someone who finds problems that can be addressed through the available data. He understands this role now more than ever.

“In this process, we were able to learn a lot about working with a lack of structure, working with a mess, and finding your way to evaluate problems and then develop a strong solution,” Brown said. “That's something you do not get to experience until something like this, and that's definitely something we can pass along to students.”

As they work to create a bridge between academia and real-world practices, Song hopes the trio will remain a team that continues to celebrate in the halls of Rawls College.

Together, he is confident they will make a difference.

“Having a good relationship with the government means a lot because we are not pure science, we are not engineering – we are the College of Business,” Song said. “The general perception outside of our college doesn't consider us science, but we are doing science as well. So, this kind of participation will enable us to create momentum in saying, ‘We are contributing to the world.'”