Tuesday, May 24, 2011

Predictive Research Aims To Improve Student Success

Web sites like Pandora, Netflix and Amazon use data collection and analysis to predict the types of music and movies a user might enjoy or to offer relevant product suggestions to online shoppers. Rio Salado College is applying similar predictive modeling technology to increase completion and success rates in the school’s online courses.

“The predictions are based on ‘habits’ in a course such as log-ins, site engagement and pace,” said Shannon Corona, Rio Salado physical science faculty chair and lead for the predictive analytics pilot program. “This is then compared to historical data against other students. We were able to predict with 70% accuracy those students who were at high-risk to not be successful in a course.”

The predictive analytics program supplements an instructor’s ability to recognize an at-risk student and helps them take steps to intervene.

“The Rio program provides instructors with an additional tool to aid students,” Corona said. “The predictive analytics program summarizes how often a student is logging into a course and using the course to be successful. “

She added, “Traditional techniques do not take this into consideration, but only look at a student's grade. An instructor now has both tools, the grade and the student engagement.”

In addition to supporting individual students, the pilot program is also being used in course design.

Rio Salado languages faculty chair Angela Felix explains, “We did not want to use anecdotal evidence. We wanted to make changes based on actual data.”

Surprisingly, grades on early assignments are not an important indicator.

“Grades are not a primary predictor of student success in an online course,” Felix said. “Log-in behavior, site engagement, and pace in the first eight days are the strongest predictors of student success.”

According to Felix, changes made to online courses in the pilot program are intended to promote the actions modeled by successful students. For example, brief introductory assignments require students to log in early and smaller more frequent assignments encourage habitual site engagement.

“We determined the factors that contribute to success in an online course and made changes to encourage these behaviors,” Felix said.

Monitoring of student engagement takes place in the background of the online learning environment.

Using methods similar to those employed by consumer Web sites like Netflix or Amazon, data is gathered and analyzed without a noticeable impact on the user experience.

Public facing elements of the program will be tested in the second summer semester of the 2011 school year. Students in pilot courses will then have the opportunity to self-monitor.

Rio Salado is one of only six institutions participating in the Predictive Analytics Reporting Framework project. PAR is a collaborative research project initiated by Western Interstate Commission for Higher Education’s Cooperative for Education Technologies and funded through a $1,000,000 grant from the Bill and Melinda Gate’s Foundation. The program will pool anonymous data collected from participating colleges to create a more accurate model of successful behavior.

Participation in cooperative programs like PAR represents a college-wide effort to improve persistence, completion and student success.