Forecasting a Better Student Experience

Predictive analytics retain students at the University of West Florida

When Dr. Jaromy Kuhl acted as dean for the Hal Marcus College of Science and Engineering at the University of West Florida in Pensacola and discovered graduation rates were only at 19 percent, he decided to take the initiative to change this number.

In collaboration with Dr. Anthony Okafor, a professor in the college’s Department of Mathematics and Statistics, Kuhl developed a new tool in 2018 to increase academic progress rates (APR) and graduation metrics in the form of a student success dashboard.

Currently dubbed the Predictive Analytics and Modeling (PAM) Lab Dashboard until an official name is patented, the tool is housed under the university’s Institute for Analytics and Industry Advancement, which is devoted to cultivating user-friendly analytics that inform data-driven decision making and predictions to improve the success of educational institutions and organizations.

“The PAM Lab Dashboard was cultivated on UWF grounds to address the needs of our students and administrators, but now we’re expanding to hopefully help universities with the same metrics as well,” said PAM Lab Director of Business Development Lesley Cox, who oversees the commercialization and expansion of the lab’s services. “We’re under NDAs [non-disclosure agreements] with a couple of universities currently, so I’m not allowed to disclose their names, but we have steady, achievable goals over the next few years to get more on board.”

The dashboard utilizes advanced predictive modeling and data visualization to personalize student success and that of the overall institution campus wide and functions to offer a comprehensive view of an institution’s health by integrating data across student demographics, course performance levels, and advisory information.

“My personal favorite feature of this tool is its ability to track at-risk students at early identification, so we can allow for timely intervention,” Cox said. “So, we track demographics from where a student went to high school and pre-college factors to on-campus statistics such as course pass rates, how often they switch majors, how many hours they take at a time, and financial aid considerations. All of that data can be used to predict and analyze future trends of how that student may perform within certain majors and programs.”

The lab uses several mathematical methods to create specific algorithms tailored to each student, Cox explained. “Much of this involves statistics machine learning and network analysis, multivariate regression, clustering classification analysis, market basket analysis (a data mining technique), and even just linear algebra,” she said.

The new algorithms expand data models to include previously unused confounding variables. For example, the consideration of whether a student took a foreign language in high school increases the predictive model’s accuracy in determining at-risk rates for first-time college students.

In addition to recording student performance, academic progress, retention, and risk levels to project prospective trends, Cox said the dashboard provides integrative tools to assist academic advisors in developing holistic action plans for students.

“Instead of being reactive, they can be proactive and have real-time interventions with students before they reach that overwhelming drop-out point,” Cox said. Too, college administrators can forecast the effect of new initiatives and potential strategies for further student support. 

While this product is the PAM lab’s pride and joy, they also offer customized data services for businesses and other institutions, such as consultant services to internal and external clients designed to assemble statistics for entities based on descriptive, diagnostic, prescriptive, predictive, and cognitive analytics. The aim of the services is to offer an examination of phenomena, how and why it occurred, how it may impact the future, and what should happen next to achieve success.

Cox said outside of education, they are primarily working in the health care sector, predicting patient outcomes with treatment plans.

“One of the things we are excited about is working with a specific health company to use historical data on treatment in cancer cells to come up with proper treatment plans,” Cox said. “We are also using a predictive model with health and behavior within sports as well as insurance cost reductions. We’re able to take everything a step further by developing industry-specific models and creating user-friendly interfaces.”

Their services provide a competitive advantage to businesses by aiding in “making data-driven decisions across the board, no matter what their background is,” Cox explained. At this writing, Cox said the PAM lab’s most current project is AI-assisted advising.

“Everyone is moving toward AI, so we have to as well,” she said. “Integrating AI will ensure we are using all our data efficiently and effectively. As it progresses and the dashboard grows, we most definitely will implement more AI on both the academic and business side of things.”

The PAM Lab measures its success through metrics. In 2020, UWF’s APR was 80.3 percent. In 2023, it increased to 86.8 percent.

“Knowing we were able to contribute to those metrics is a reward in itself for us,” Cox said. “And, I think it makes the [UWF]president really happy.”