No Length Left Behind: Enhancing Knowledge Tracing for Modeling Sequences of Excessive or Insufficient Lengths

TL;DR

This paper introduces SFKT, a novel knowledge tracing model designed to effectively handle sequences of varying lengths, especially very long or very short, improving the modeling of student learning behaviors.

Contribution

The paper proposes SFKT, a new KT model that adaptively manages sequences of different lengths, overcoming limitations of truncation and overfitting in existing methods.

Findings

SFKT improves prediction accuracy over baseline models.

SFKT effectively models both long and short student sequences.

The approach reduces computational costs for long sequences.

Abstract

Knowledge tracing (KT) aims to predict students' responses to practices based on their historical question-answering behaviors. However, most current KT methods focus on improving overall AUC, leaving ample room for optimization in modeling sequences of excessive or insufficient lengths. As sequences get longer, computational costs will increase exponentially. Therefore, KT methods usually truncate sequences to an acceptable length, which makes it difficult for models on online service systems to capture complete historical practice behaviors of students with too long sequences. Conversely, modeling students with short practice sequences using most KT methods may result in overfitting due to limited observation samples. To address the above limitations, we propose a model called Sequence-Flexible Knowledge Tracing (SFKT).