The Opportunity Count Model: A Flexible Approach to Modeling Student Performance
Rich features can be exploited to better model student performance when predicting next problem correctness (NPC) within intelligent tutoring systems. Yet these features may differ significantly in availability and importance when considering opportunity count (OC), or the number of problems experienced within a skill or knowledge component. Inspired by such intuition, the present study examines the Opportunity Count Model (OCM), a unique approach to student modeling in which separate models are built for differing OCs rather than creating a blanket model to encompass all OCs. Random Forest (RF) is used to establish iterations of the OCM by considering rich features within logged tutor data. Model strength is then tested against standard Knowledge Tracing. Results suggest that prediction of next problem correctness is improved through the OCM approach for lower OCs, and applying different modeling techniques at different phase of students’ practice would be plausible. Also, feature variation among OCs justifies our proposal to build OCM.
Code and data set used in the paper can be found here. Please note that code will be better organized soon.