Description

In this project, I explored how additional variables can improve the accuracy of Deep Knowledge Tracing (DKT) in predicting student performance. I compared DKT with the traditional Bayesian Knowledge Tracing (BKT) on a dataset of adult learners from the Workera platform. By incorporating variables like time to answer, skill, and sub-skill, I demonstrated that these additions significantly enhance the performance of DKT. This research provides valuable insights for intelligent tutoring systems, enabling better prediction of student outcomes and more tailored educational experiences.

Scientific paper

Paper presentation video

Github Repo