-->

Winning Texas A&M Statistics team members (starting second from left) Bryce Durgin, Kun Xu, (starting second from far right) Stephanie Whang and Ranye Sun pose with Texas A&M professors Simon Sheather and Edward Jones (center), along with Scott Hallworth, Capital One Senior Vice President of Risk Management (far left). (Photos courtesy of Texas A&M Statistics)

COLLEGE STATION --

The challenge issued by banking giant Capital One was complex: Develop a statistical model to identify fraudulent credit-card transactions within 400,000 records that each contain 340 variables.

In the end, a panel of the company's judges selected a group of graduate students from the Texas A&M University Department of Statistics as the winning team among 28 that participated in weeks-long competition. The students -- Bryce Durgin, Ranye Sun, Stephanie Whang and Kun Xu -- each received a $1,000 cash prize.

As the final stage of a grueling competition that began in October, Capital One selected five finalist teams to make 20-minute presentations on Nov. 29 before judges at the company's international headquarters in McLean, Va. In addition to the winning team, Texas A&M also fielded a second team that advanced to the final five: Yichen Chang, Ming Lu, Jingang Miao, Ya Su and Rubin Wei. All finalists had their travel expenses paid by Capital One, and each received a new Kindle Fire as a reward for being members of the top five teams.

Simon Sheather, professor and head of Texas A&M Statistics, said the competition is designed to help Capital One identify and recruit top-talent statisticians -- the very type of quality leaders his department has a proven track record for developing. Texas A&M is ranked No. 3 in the nation among public graduate programs in statistics and biostatistics according to the latest U.S. News and World Report rankings, and Sheather credits in part the program's emphasis on real-world experience. All students, for instance, work on a real consulting project before graduation -- preparation that obviously paid off in Virginia.

"It was an amazing accomplishment that reinforces the fact that we have a long tradition of training people well for both industry and academia," Sheather said.

Edward Jones, an executive professor of statistics who recruited and mentored the Texas A&M teams' members, said the winning team used the discovery of new variables to partition the massive data set into 11 segments. They then employed the statistical concepts of logistic regression and stepwise regression to develop the final models for forecasting the probability that a transaction was fraudulent. They also had to factor in the costs associated with both rejecting a non-fraudulent transaction and not rejecting a fraudulent one.

"The real strength of their solution was that it was not only clever, it was easy to describe the relative importance of key variables to executives and marketing experts," Jones said.

In addition to two of the five finalists, Texas A&M also had a third team in the competition that consisted of distance learning students who did not have the luxury of meeting face-to-face. Team members included Justin Bein, Steven Jackson, Yonatan Negash and Samuel Temple.

Sheather said he is proud of each of the students who participated in the complex work that consumed roughly 20-to-30 hours a week of the students' lives for several weeks.

"This was amazing work," Sheather said. "As somebody who has 30 years of statistical modeling experience, this was a difficult project. Even a faculty member would have found the model associated with this challenging."

For additional information about Texas A&M Statistics, visit http://www.stat.tamu.edu/.

-aTm-

Contact: Vimal Patel, (979) 845-7246 or vpatel@science.tamu.edu or Dr. Simon Sheather, (979) 845-3141 or sheather@tamu.edu

Patel Vimal

  • Another Fab Five

    Texas A&M's second team advancing to the final top five featured (from left) Yichen Chang, Ming Lu, Jingang Miao, Ya Su and Rubin Wei.

  • Stand and Deliver

    Ranye Sun, presenting for Texas A&M's first-place team.

  • Finding Fraud

    The competition, designed by Capital One to help identify the top statistical performers and potential future employees, charged students with developing statistical models to identify fraudulent credit-card transactions within 400,000 records, each of which contained 340 variables.

ALL RIGHTS RESERVED
© Texas A&M University. To request use of any of our photographs for educational use or to view additional options from our archive, please contact the College of Science Communications Office.

College of Science
517 Blocker
TAMU 3257 | 979-845-7361
Site Policies
Contact Webmaster
Social Media