qDKT: Question-centric Deep Knowledge Tracing

TL;DR

qDKT is a novel question-centric knowledge tracing model that improves prediction accuracy by modeling individual questions and applying graph Laplacian regularization, outperforming existing models on real-world datasets.

Contribution

It introduces qDKT, a question-centric KT model with graph Laplacian regularization and an innovative initialization scheme, advancing the accuracy of learner performance predictions.

Findings

qDKT achieves state-of-the-art results on multiple datasets.

The model effectively captures question-specific information.

Regularization improves predictions in large question sets.

Abstract

Knowledge tracing (KT) models, e.g., the deep knowledge tracing (DKT) model, track an individual learner's acquisition of skills over time by examining the learner's performance on questions related to those skills. A practical limitation in most existing KT models is that all questions nested under a particular skill are treated as equivalent observations of a learner's ability, which is an inaccurate assumption in real-world educational scenarios. To overcome this limitation we introduce qDKT, a variant of DKT that models every learner's success probability on individual questions over time. First, qDKT incorporates graph Laplacian regularization to smooth predictions under each skill, which is particularly useful when the number of questions in the dataset is big. Second, qDKT uses an initialization scheme inspired by the fastText algorithm, which has found success in a variety of language modeling tasks. Our experiments on several real-world datasets show that qDKT achieves state-of-art performance on predicting learner outcomes. Because of this, qDKT can serve as a simple, yet tough-to-beat, baseline for new question-centric KT models.