Industry Partners Invest in Statistics Students
UCF’s winning team: Yanmei Patella, Jingrong Dai, Mingming Zhou; facutly advisor: Daoji Li, Ph.D.; CFE’s team: Jason Mizrahi, Kristen Ward and Daniel Kenon
From developing predictive models to interpreting data, UCF graduate students had the chance to put their data analytics skills to the test, as CFE Federal Credit Union launched its inaugural Lending Analytics Competition with the UCF Department of Statistics.
From October 2017 to January 2018, student teams used data sets to predict lending risks facing the credit union industry.
“As the field of data analytics grows exponentially, it has become more important than ever for companies to use big data to make strategic decisions,” said Kevin Miller, CFE’s President and CEO. “At CFE, we believe in the value of data analytics in the banking industry and beyond, and are proud to invest in UCF’s graduate-level programs.”
Together CFE’s Business Intelligence department and the UCF Department of Statistics developed a competition designed to reduce consumer lending risks. Teams received data sets with disguised information about CFE’s loan portfolio, along with data dictionaries to help students better understand the industry-specific data. With this information, teams created predictive models to determine which data points are most closely associated with loan charge-offs, which is when a person fails to repay their loan. The student teams communicated their findings in a project report to present to senior management.
The first place team included Mingming Zhou, Jingrong Dai and Yanmei Patella, who worked almost every day during the competition period and used their analytic skills, data mining knowledge and the SAS platform to accurately predict whether repaying a loan would fail, and what the expected loss would be for CFE.
“Our students work very hard and produce incredible head-turning work,” said Assistant Professor, Daoji Li, Ph.D., who advised the winning team. “Winning this competition is a very noteworthy accomplishment because it is not a toy project—it’s a real one.”
Two students from the winning team took Li’s Data Mining II class during the spring 2017 semester. He also teaches Data Mining I in the fall semester of every year.
“This win makes me happiest because we can use the data mining and analytical skills learned from Dr. Li’s class to successfully solve a real-world problem,” said Zhou, the captain of the winning team. “I hope to be a data scientist after graduation from UCF. This is a good experience.”
Zhou would know – she was also on the team who finished in second place for the 2017 SAS Analytics Shootout competition, also sponsored by Li.
Graduate student Michael Koller was on the team that took third place. He and his teammates worked on the data set for months, trying to find different factors that could help CFE predict risk associated with lending loans. This was Koller’s first time working on a such a large project.
“The biggest thing I learned was how long it takes to get a data set in industry to get to something usable that the people can really understand,” Koller said. “We had a subset of the portfolio that was 20,000 different approved loans.”
The teams’ models were tested against a wider data set, and three teams were selected based on how closely their models predicted factors related to failure of loan repayment.
“Our main finding was that if the loan borrower had a pre-existing relationship with CFE they had a much lower risk of failing,” Koller said.
On March 22, UCF hosted its annual Data Analytics Symposium, featuring professors and industry professionals as keynote speakers. During the event, CFE announced the three teams, and the first-place winners had the opportunity to present their findings. CFE awarded a total of $5,000 in scholarships to the winning teams.
“By working with talented students at UCF, we can resolve real-world problems that are facing our organization and our industry as a whole. Through data analytics, students have the power to effect real change in the organizations they serve, as they collaborate with businesses both locally and worldwide,” said Kevin Wright, Chief Information Officer at CFE.
CFE has partnered with the Department of Statistics in the past to support their programs and hire graduate-level interns. In the future, CFE plans to host additional data analytics competitions that address issues facing the banking industry.
“We look forward to more CFE-sponsored competition opportunities for our students in the future and more collaborations, especially with our Ph.D. program in Big Data,” Statistics Chair Shunpu Zhang, Ph.D. said. “The students really learned the difference between textbook data and real data and practiced their skills. The competition was a success.”