Like many universities, Purdue has been working on predictive analytics for quite some time. But until now, much of that information hasn't made it into students' hands.
Enter Forecast, a student-facing Web application designed to compare current student behavior with the aggregated behavior of successful peers over the last eight years. Over the past year, the Indiana university adopted hardware and software platforms to support predictive analytics, and then developed the application for students on top of that infrastructure. Purdue also plans to purchase an application for advisers to pilot in the spring that will take advantage of the student behavior data.
"We really wanted to expand what we were doing in the space of predictive modeling to allow us to look at more near-time and real-time data points of how students are actually interacting with the campus, and see how those were or were not predictive of whether they were ultimately successful here," said Brent M. Drake, who became Purdue's first chief data officer nearly three years ago.
Those data points include student grades, progress toward a degree, use of the course management system, wireless network use and swipes of their Purdue ID card on campus locations. All of this data is being integrated from four major sources on campus: the student information system, learning management system, network logs and 70 university ID card transaction databases.
The public component of the application already has three key behaviors that encourage student success: Don't wait to add classes, get engaged in campus activities and compare your GPA with your peers. Next to each of these suggested behaviors, the university included resources that will help them reach those goals. For example, students who want to get a higher GPA can learn about good time management practices and tutoring options.
When students log into the application, they see how their results on these behaviors match up to aggregated results of successful students. The university will send them emails and website updates about their progress, and eventually, Drake would like to have a mobile app developed that will share push notifications with students.
While some of this data is rather traditional, it's a bit unusual to use network logs and university ID card swipes in predictive modeling. Purdue is trying to find a way to measure student engagement on campus outside of classes, which could correlate to students' GPA. Admittedly, Drake said that measuring the times that student devices connect on the network and when they go in a campus building doesn't directly translate to student engagement academically on campus. It's possible that students could lie out in the quad for eight hours and not do anything academically to help them learn. But these two behaviors could give them some insight.
"We use the data as a nudge to say, 'Here's a reasonable proxy of how much it appears you're engaged on campus, and here's a list of things that can help you do it in a more productive manner,'" said Drake.
Similarly adding classes after the semester has already started may be an indication that the students will have a harder time academically. But that also is not a cause-and-effect relationship, he said.
Purdue plans to study whether this application will help students actually change their behavior and improve academically. In the meantime, it's hard at work on a student success collaborative initiative that will trigger alerts to advisers when students don't register in time for a class, aren't engaged on campus or drop a class, among other things. Together, the university would like to see these systems help students grow academically.
"We have, I believe, a very strong commitment to helping our students succeed and using our campus data to make that a reality," Drake said. "And I'm very encouraged by the steps we're doing in this space to try to make that more of a reality."