Probability based scaffolding system using Deep Learning

Abstract

Recently, a great deal of interest in the learning science field has arisen in the use of software to scaffold students in complex tasks. However, most of those software tools have been unable to adapt to individuals To solve the problem, IRT-based approaches to predict student's performance have been proposed. These studies show predicting students' correct answer probability with high accuracy is of critical importance. However, IRT-based approach doesn't predict student's performance accurately when the test data are sparse or imbalanced. To achieve high accuracy in those situations, we proposed a novel scaffolding system based on deep learning. We show proposed method can predict student's performance more precisely than traditional IRT method.