Personalized learning using graph neural network and knowledge graph

Abstract

Personalized learning is designed to serve as a customization tool for providing recommendations. Some examples are recommending study pathways and suggesting suitable next questions when students’ knowledge have been tracked/traced. Deep Knowledge Tracing (DKT) is the first deep learning model that traces the knowledge of students using the long short-term memory (LSTM) model. In this thesis, we propose to exploit the Graph Neural Network (GNN) model and the knowledge graph in the DKT model to tackle the limitations of the traditional DKT model without considering the knowledge graph structure. We demonstrate how the graph model can be used to improve the DKT model with the help of the knowledge graph structure. Our model can be applied to modern e-learning systems for personalized learning which predicts the future performance of students and recommends questions which are suitable to students for improvement.

Date of Award	2020
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Raymond Chi Wing WONG (Supervisor)