Personalised Adaptive E-Learning System with Competency Graph and Machine Learning

Abstract

Adaptive e-learning systems are designed to tailor learning materials based on a learner's performance and mastery progression. However, existing systems use static unlocking mechanisms based on rules and do not provide a structured representation of topic dependencies. In this paper, we propose a personalised adaptive e-learning system that incorporates a competency graph with machine learning-based mastery prediction to dynamically shift quiz difficulty and recommend the best learning paths. The system represents domain knowledge as a directed graph of prerequisite-order-based competencies and monitors learner interaction data, such as quiz scores, attempts, and time spent. A model is trained to predict the correct mastery and difficulty, which it can dynamically adjust. We also compare our approaches to existing adaptive models and demonstrate improved structural personalisation and mastery-driven progression compared to traditional rule-based systems. The framework presented features such as scalability, enforced structured learning, and improved learner engagement compared to traditional e-learning systems.