Skip to main content Skip to navigation

Enhancing Deep Knowledge Tracing via Diffusion Models for Personalized Adaptive Learning

Project Overview

The document explores the application of generative AI, particularly diffusion models, to improve Deep Knowledge Tracing (DKT) within the framework of Personalized Adaptive Learning (PAL). By generating synthetic educational data, this approach effectively tackles the issue of data scarcity in student learning records, which is crucial for accurate modeling and prediction of student performance. The findings reveal that incorporating AI-generated data leads to substantial enhancements in DKT outcomes, especially in scenarios where training data is limited. This innovation not only supports the personalization of learning experiences but also demonstrates the potential of generative AI to transform educational practices by providing robust data solutions that facilitate better learning analytics and student engagement.

Key Applications

TabDDPM (a diffusion model for generating synthetic educational records)

Context: Personalized Adaptive Learning (PAL) in educational settings, targeting educators and students.

Implementation: The method involves sampling existing student data to create synthetic records using TabDDPM, which are then combined with original data to train DKT models.

Outcomes: Significant improvement in DKT performance metrics, including accuracy, AUC, precision, and recall, particularly in low-data training scenarios.

Challenges: Limited availability of real educational data for training DKT models, necessitating synthetic data generation.

Implementation Barriers

Data Availability

Scarcity of training data for effective knowledge tracing, which limits model performance.

Proposed Solutions: Utilizing generative AI models like TabDDPM to create synthetic educational records to augment training datasets.

Project Team

Ming Kuo

Researcher

Shouvon Sarker

Researcher

Lijun Qian

Researcher

Yujian Fu

Researcher

Xiangfang Li

Researcher

Xishuang Dong

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Ming Kuo, Shouvon Sarker, Lijun Qian, Yujian Fu, Xiangfang Li, Xishuang Dong

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies