MBA student uses AI to perfect class selection process

Derek Gibbs, an MBA student at the Wharton School and vice president of innovation of the Wharton AI & Analytics Club, has created a prediction algorithm that, with assistance from AI, helps MBA students more accurately forecast their class schedules.

Wharton utilizes a unique course scheduling system called CourseMatch in which students express their preferences (called “utilities” in the MBA program) for a class on a 0-100 scale. If a student is really excited about a class, they might score it 100; if they’re marginally interested in a class, they might mark 50; if they are unenthused about a class but willing to take it, they might put 1 or 5.

However, Gibbs says students often struggle converting their preferences to numbers and have no idea what classes and overall class schedule they will receive. They could get all, one, or none of them.

Gibbs’ prediction algorithm, which he is calling CourseCast, solves this problem by informing students what classes, on average, they’re likely to receive from the CourseMatch system.

“It gives you the likelihood of getting specific classes,” he says. “It’ll say you have a 90% chance of getting class A, an 80% change of getting class B, and the same is true for entire schedules. It’ll say you have a 50% chance of getting classes A, C, E, and F.”

The tool uses historical data to estimate the likelihood of receiving specific courses based on the students’ expressed utilities. If they are unhappy with their forecasted schedule, they can revise and resubmit their preferences to try to get a more suitable result.

Gibbs, who studied finance and information systems while an undergrad at Temple, says he spent around 100 hours across three semesters working on CourseCast, starting with a simple spreadsheet and then a prototype, before designing the final product. He used AI to do the actual coding, using tools such as Cursor and Claude. He says Wharton was very supportive along the way, readily providing him with the data he needed to build the forecasting tool.

Gérard Cachon, a professor of marketing and a professor of operations, information, and decisions at the Wharton School, served as Gibbs’ adviser to an independent study he did to explore the prediction algorithm concept. Gibbs says he worked closely with Cachon—one of the original professors who invented CourseMatch in 2014—to get a better understanding of how the system works.

Cachon used the old adage “garbage out, garbage in” to describe how CourseCast improves upon the CourseMatch system.

“If students can’t accurately report their preferences to CourseMatch, then CourseMatch can optimize all it wants, but students will not be happy,” he says, “so it is critical that students are able to report their preferences accurately, and CourseCast is a wonderful tool to help them do that.”

When Gibbs approached him with the idea, Cachon said he liked it, and says the final product is something Wharton can be very proud of. 

“Students are happier with the process of getting their course schedule,” he says. “They spend less time doing it and they get courses that better match their preferences.”

Last fall, Gibbs and the Wharton AI & Analytics Club ran a competition to see if members could improve on his model before it launched. Gibbs say while no single student’s model was the one that they ended up using, a lot of the learnings from the hackathon were rolled in to the final prediction model.

Since its official launch last year, Gibbs says more than 1,200 out of Wharton’s 1,600 MBA students have used CourseCast, and it has received positive reviews.

“Everyone that I’ve talked to and through the surveys typically received either exactly the schedule that they were forecasted to receive, or something very, very close,” he says.

Gibbs says he is still thinking through ways to improve the forecasting and is currently in talks with the Wharton administration to try to make CourseCast a permanent tool.

“My goal would be for no one to get a schedule that they don’t want,” he says. “Obviously not everyone can get into every single really popular class, but I think there is a schedule out there for every student that gives them a little bit of what they want and they’re happy with it.” 

Artificial intelligence touches disciplines across campus. In a limited spring profile series, Penn Today is highlighting innovative students at Penn who are adopting this technology in a variety of projects. To learn more about how members of the Penn community are pioneering the understanding and advancement of AI, visit the Penn AI website.