Dynamic Knowledge Tracing Models for Large-Scale Adaptive Learning Environments

— Large-scale data about learners’ behavior are being generated at high speed on various online learning platforms. Knowledge Tracing (KT) is a family of machine learning sequence models that use these data to identify the likelihood of future learning performance. KT models hold great potential for the online education industry by enabling the development of personalized adaptive learning systems. This study provides an overview of five KT models from both a technical and an educational point of view. Each model is chosen based on the inclusion of at least one adaptive learning property. These are the recency effects of engagement with the learning resources, dynamic sequences of learning resources, inclusion of students’ differences, and learning resources dependencies. Furthermore, the study outlines for each model, the data representation, evaluation, and optimization component, together with their advantages and potential pitfalls. The aforementioned dimensions and the underlying model assumptions reveal potential strengths and weaknesses of each model with regard to a specific application. Based on the need for advanced analytical methods suited for large-scale data, we briefly review big data analytics along with KT learning algorithms’ scalability. Challenges and future research directions regarding learners’ performance prediction are outlined. The provided overview is intended to serve as a guide for researchers and system developers, linking the models to the learner’s knowledge acquisition process modeled over time.