Learning Behavior Analytics for Adaptive Path Optimization in Online Education Systems

Abstract

The growing multiplicity of learner behaviour in online education systems has opened up the constraints of fixed learning paths that do not move with the individual engagement, performance and the advanced knowledge. Although the available data driven methods provide various degrees of personalization, most of them are based on either heuristic decision rules or black-box machine learning models that are not optimised and interpretable. The current paper introduces a behaviour-sensitive behaviour adaptive learning path optimization framework, which combines learning behaviour analytics and a sequential optimization formulation. The dynamics between learners are modelled as a multidimensional state representation of dynamics of engagement, mastery and performance. According to this model, adaptive learning path selection is transformed into a sequence of decision-making processes, and a behaviour-conscious optimization algorithm is created, which can be used to dynamically determine a learning activity that leads to the greatest cumulative learning increase. The so-called framework is assessed based on an online learning dataset and compared against a set of the methods that are based on the static and machine learning. The experiments prove increased learning outcome and enhanced course completion of the course, which proves the appropriateness of behaviour-based adaptive optimization. These results demonstrate that behaviour modelling in terms of formalisation and adaptive optimization can be employed to a large extent to personalise and increase the effectiveness of learning in online education settings.