home - login - register

PERC 2012 Abstract Detail Page

Previous Page  |  New Search  |  Browse All

Abstract Title: Multidimensional Student Skills from Collaborative Filtering
Abstract: Despite the fact that a physics course typically culminates in one final grade for the student, many instructors and researchers believe that there are several different skills that students need to acquire to achieve mastery. Reasonably large data sets, from online assignments in blended courses or from widely administered standard instruments, lend themselves to analysis using a statistical learning algorithm called Collaborative Filtering (CF), which can be applied to discover the "dimensionality" of the skills involved and the clustering of assessment items by their ability to discriminate between students who have more or less of these skills.  CF in this context is related to factor analysis and item response theory (IRT), as we show, but it comes at the problem from a machine learning perspective, seeking to maximize the accuracy in predicting which students will answer which items correctly. We describe the operation of a CF and demonstrate its application to student response data in coursework and on standard instruments.
Abstract Type: Contributed Poster Presentation

Author/Organizer Information

Primary Contact: Yoav Bergner
Massachusetts Institute of Technology
Co-Author(s)
and Co-Presenter(s)
Stefan Droschler, Gerd Kortemeyer, Saif Rayyan, Daniel Seaton, David Pritchard