Abstract:The goal of Knowledge Tracing (KT) is to estimate how well students have mastered a concept based on their historical learning of related exercises. The benefit of knowledge tracing is that students' learning plans can be better organised and adjusted, and interventions can be made when necessary. With the recent rise of deep learning, Deep Knowledge Tracing (DKT) has utilised Recurrent Neural Networks (RNNs) to accomplish this task with some success. Other works have attempted to introduce Graph Neural Networks (GNNs) and redefine the task accordingly to achieve significant improvements. However, these efforts suffer from at least one of the following drawbacks: (1) they pay too much attention to details of the nodes rather than to high-level semantic information; (2) they struggle to effectively establish spatial associations and complex structures of the nodes; and (3) they represent either concepts or exercises only, without integrating them. Inspired by recent advances in self-supervised learning, we propose a Bi-Graph Contrastive Learning based Knowledge Tracing (Bi-CLKT) to address these limitations. Specifically, we design a two-layer comparative learning scheme based on an "exercise-to-exercise" (E2E) relational subgraph. It involves node-level contrastive learning of subgraphs to obtain discriminative representations of exercises, and graph-level contrastive learning to obtain discriminative representations of concepts. Moreover, we designed a joint contrastive loss to obtain better representations and hence better prediction performance. Also, we explored two different variants, using RNN and memory-augmented neural networks as the prediction layer for comparison to obtain better representations of exercises and concepts respectively. Extensive experiments on four real-world datasets show that the proposed Bi-CLKT and its variants outperform other baseline models.

GMEKT: A Novel Graph Attention-Based Memory-Enhanced Knowledge Tracing.

A Temporal Convolutional Knowledge Tracing Model Integrating Forgetting Factors and Item Response Theory

Graph Attention Neural Network Model with Behavior Features for Knowledge Tracking.

Knowledge Relation Rank Enhanced Heterogeneous Learning Interaction Modeling for Neural Graph Forgetting Knowledge Tracing

Deep Graph Memory Networks for Forgetting-Robust Knowledge Tracing

A Deeper Knowledge Tracking Model Integrating Cognitive Theory and Learning Behavior

SQKT: A Student Attention-Based and Question-Aware Model for Knowledge Tracing

DAGKT: Difficulty and Attempts Boosted Graph-based Knowledge Tracing

MAN: Memory-augmented Attentive Networks for Deep Learning-based Knowledge Tracing

DGEKT: A Dual Graph Ensemble Learning Method for Knowledge Tracing

GIKT: A Graph-Based Interaction Model for Knowledge Tracing

GKT-CD: Make Cognitive Diagnosis Model Enhanced by Graph-based Knowledge Tracing

ELAKT: Enhancing Locality for Attentive Knowledge Tracing

Automatical Graph-based Knowledge Tracing

DyGKT: Dynamic Graph Learning for Knowledge Tracing

Augmenting Interpretable Knowledge Tracing by Ability Attribute and Attention Mechanism

Temporal Graph Memory Networks For Knowledge Tracing

Integrating fine-grained attention into multi-task learning for knowledge tracing

Bi-CLKT: Bi-Graph Contrastive Learning based Knowledge Tracing

Modeling Student Performance Using Feature Crosses Information for Knowledge Tracing

Knowledge Tracing with Sequential Key-Value Memory Networks