Continual learning in knowledge tracing

• Main Author: Sujanya, Suresh
• Other Authors: Ponnuthurai Nagaratnam Suganthan
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/157960
• Institution: Nanyang Technological University

The key to building a more sustainable world is high-quality education. The recent COVID-19 pandemic has sparked a surge in online education, allowing students and teachers to learn and teach from the comfort of their own homes.This has led to large amount of student learning activities data being collected. Knowledge Tracing (KT), which aims to monitor learners’ evolving knowledge state and evaluate their growing knowledge acquisitions, is a crucial and vital component in online learning. The learning assessments depend on the ability of a student to learn and master a skill based on the history of their performance. However, due to data privacy concerns, it is difficult to combine the learners’ data from multiple schools, and the learning of newer tasks leads to forgetting of the older ones. Hence, this work explores the feasibility of developing these models while preserving the confidentiality of learners’ data and customizing the learning experiences within their schools. This study is conducted using a portion of the ASSISTments dataset (2009) in a continual learning framework adapting the Self Attentive Knowledge Tracing (SAKT) algorithm. The outcomes achieved by learning sequentially in a task-incremental setting are better than pooling all the data together. Keywords: Knowledge Tracing, Continual learning, catastrophic forgetting.