“Some of my faculty colleagues were unconvinced of the usefulness of predictive analytics, fearing the standard error would be too large. I had a different mindset. Without further testing how could we know?”
– Dr. Jonathan Hilpert, Professor, Georgia Southern University
With only 56% of students in the US completing their bachelor’s degree in six years, student success is a hot button topic for educators, students, and government agencies. It is also one of the most challenging issues facing higher education today.
Educational institutions need to better understand what factors play a significant role in a student’s learning success. In order to get a complete picture of a student’s progress in real time, they need to turn to the data sitting within their learning management system (LMS) and other IT support systems and put it to use.
Georgia Southern University (Georgia Southern) shares this sentiment. And to prove it, they decided to pilot a new predictive analytics solution to better identify at-risk students and help them succeed. The university needed the help of faculty to test it out, but not every faculty member was convinced of the accuracy of predictive analytics.
Dr. Jonathan Hilpert, an Associate Professor within Georgia Southern’s College of Education disagreed. He knew that piloting a solution was necessary to make an educated decision, so he became the first to allow data from his courses to be analyzed using predictive analytics to test their accuracy – and the payoff was worth it.
Beginning the Predictive Analytics Journey
Hilpert teaches in graduate programs where most of the courses are delivered online, with no face-to-face access to students. With five semesters of student data available, Georgia Southern’s IT team had a solid baseline of data for the pilot, providing Hilpert with predictive analytics for his current course.
Making Sense of the Data
Georgia Southern’s IT team pulled together data from five of Hilpert’s classes, over five semesters, to feed into the solution. This would allow the system to track how a student performed in a virtual environment, how they scored, and then use those patterns to predict how they would perform going forward. After an analysis was run on courses from the previous semester, it showed the new solution predicting student outcomes with 70% accuracy, just 10 weeks into the course.*
Despite a strong start, this student is considered at risk. Major factors include not participating in course discussions and performing poorly on assessments.
“It was interesting to see with a certain degree of confidence how a student is performing. Analytics are predictors of latent factors and unobservable influences that, can help communicate different messages to students to improve their success.”- Dr. Jonathan Hilpert, Professor, Georgia Southern University
How to Avoid Drowning in Data
With several thousand undergraduate students flowing through Georgia Southern programs, Hilpert also sees how one could begin to drown in the volume of data coming in. He suggests focusing the power of predictive analytics on student populations already known to be “at-risk.”
At-Risk Student widget gives the professor a quick view of students at-risk and the main contributing factors.
“Within our College, we have a group of graduate students who are conditionally admitted into the program. If they make a C in the introductory course, they can’t continue with the program,” he explains. “So every year, we lose a handful of conditional grad students. If we had predictive analytics providing better feedback to instructors, we may be able to more quickly identify those conditional admits who are struggling, and advocate for a certain type of engagement or message to help them overcome their challenges.”
Doing More with Data
Hilpert’s exposure to a solution that can effectively support predictive analytics has left him wanting more.
“The system won’t make decisions for you, but will help you make decisions. If I had access to predictive analytics day-to-day in the classroom, I’d probably use it. We’re talking about real-time, data-driven decision-making, which in an online teaching environment is incredibly useful. A lot of the time, our students have never taken an online course before. They need some kind of indication in how they are doing or performing. If a computer can monitor the situation and pick up more quickly that a student is trending down, then an instructor can use that information to get a dialogue going with the student. And that’s beneficial to everyone.”
If you want to learn how Georgia Southern University used their learner data to improve retention and graduation rates, you need to watch this webinar. They’re able to identify at-risk students and intervene before it’s too late. As early as the fourth week of classes, their success index was able to predict student outcomes to within a letter grade 67% of the time*.