Implementation of a Data-Driven Solution for Student Loans: Utilizing Data Mining Algorithms Approach
Session Title
Accounting, Finance, and Economics
College
College of Business Administration
Department
Accounting, Finance & Economics
Abstract
As of 2020, student loans debt hit $1.6 trillion, with private loan debt volume reaching over $125 billion. Student loans have grown to become the second largest category of household debt in the U.S. It has also become the largest financial burden in terms of debt for graduates, with nearly 44 million individuals holding outstanding student loans. The private loan industry accounts for about 8% of the market. The private sector, even with a stronger underwriting process, still has a relatively high default rate with about 1 in 10 individuals defaulting on their loans. Credit-risk assessments conducted by these private lending institutions are heavily reliant on variables such as debt-to-income ratio, credit history, FICO scores and co-signer availability. This paper explores a data mining algorithmic approach with the utilization of “untraditional” variables to determine an individual’s credit risk in regard to student loans. Using a neural network model with data from the U.S Department of Education, the aim is to extrapolate a reliable predictive model affecting student loan repayment. The goal is also to understand the business viability and business integration value of an automated credit-risk assessment tool that in theory should reduce default risk and increase efficiency by eliminating one of lending institutions’ major areas of overhead: underwriting costs.
Honors Thesis Committee
Yuanshan Cheng, Ph.D.; Philip Gibson, Ph.D.; and Shirley Shen, Ph.D.
Start Date
24-4-2020 12:00 AM
Implementation of a Data-Driven Solution for Student Loans: Utilizing Data Mining Algorithms Approach
As of 2020, student loans debt hit $1.6 trillion, with private loan debt volume reaching over $125 billion. Student loans have grown to become the second largest category of household debt in the U.S. It has also become the largest financial burden in terms of debt for graduates, with nearly 44 million individuals holding outstanding student loans. The private loan industry accounts for about 8% of the market. The private sector, even with a stronger underwriting process, still has a relatively high default rate with about 1 in 10 individuals defaulting on their loans. Credit-risk assessments conducted by these private lending institutions are heavily reliant on variables such as debt-to-income ratio, credit history, FICO scores and co-signer availability. This paper explores a data mining algorithmic approach with the utilization of “untraditional” variables to determine an individual’s credit risk in regard to student loans. Using a neural network model with data from the U.S Department of Education, the aim is to extrapolate a reliable predictive model affecting student loan repayment. The goal is also to understand the business viability and business integration value of an automated credit-risk assessment tool that in theory should reduce default risk and increase efficiency by eliminating one of lending institutions’ major areas of overhead: underwriting costs.