Netflix-Like Algorithm Drives New College-Finding Tool

As an admissions counselor at Valparaiso University, Daniel Jarratt noticed that few high-school students really knew what they were looking for in a college. For all the talk about the importance of college choice, most students Mr. Jarratt spoke to knew of a few colleges they wanted to attend but couldn’t articulate exactly why they wanted to do so.

So on his nights and weekends, Mr. Jarratt, now a first-year Ph.D. candidate in computer science at the University of Minnesota-Twin Cities, started working on a tool that would direct students to the right colleges even if they didn’t know what they were looking for.

He began by downloading data from the U.S. Department of Education’s Integrated Postsecondary Education Data System, or Ipeds, and then researched the science of college recommendation to figure out which data points he should use to compare institutions. He settled on 80 variables that together determine the feel of a college—everything from the number of National Merit Scholars to the types of majors.

Mr. Jarratt then created an algorithm that could take several colleges and figure out how similar they are to one another and—more important—in what ways they are similar. Do they have a lower-than-average graduation rate? More than the average number of students living on the campus? Do a higher-than-average number of students study art or engineering?

Using that algorithm, he could explain what the students could not: what was it that a collection of colleges had in common. From there, Mr. Jarratt could highlight other institutions that shared some of the same attributes.

Mr. Jarratt’s algorithm is now an integral part of PossibilityU, a website that helps high-school students find the right college.

PossibilityU’s data-driven approach to college matching isn’t new, but Mr. Jarratt’s recommendation algorithm is unique. Rather than starting with a list of questions about what students are looking for, PossibilityU asks users to enter up to three colleges that they are interested in. It then spits out a list of 10 other, similar colleges to consider. A premium paid subscription allows students to compare an unlimited number of colleges and provides application deadlines and other advice.

It’s kind of like Netflix’s movie suggestions, says Mr. Jarratt, who studies recommender systems like those used by the movie service and by Amazon.

Do you like Valparaiso and the University of Minnesota? You might also like Marquette University and the University of Iowa, according to PossibilityU. The tool then tells you that those matches are based on a number of characteristics:

  • Similar geographic location.
  • Significantly higher-than-average graduation rate.
  • Significantly higher-than-average number of majors and advanced degrees offered.
  • Slightly higher-than-average percentage of students who live on the campus.

“We’re using the characteristics of colleges to get at the nascent preferences of students,” Mr. Jarratt says.

Link to Full Article: http://chronicle.com/blogs/data/2014/01/23/netflix-like-algorithm-drives-new-college-finding-tool/