Scholarlink

<p>As digital learning environments continue to reshape higher education, understanding the stability and dynamics of student engagement is essential for optimizing platform design and instructional outcomes. This study examines the behavioral evolution of students utilizing a Moodle-based e-learning system and evaluates the effect of targeted system improvements on platform engagement and learning outcomes. To achieve this, we employ a hybrid analytical framework that integrates interpretable machine learning, namely Individual Conditional Expectation (ICE) and Partial Dependence Plots (PDP), with chaos-theoretic indicators, including Lyapunov Exponent, Fractal Dimension, Butterfly Effect, and Hurst Exponent. Using data collected from 500 university students over a 28-week period, we examine the behavioral changes that followed the introduction of a system improvement introduced at Week 11. The results reveal a substantial shift in engagement patterns following the intervention, including increased login frequency, extended session duration, and higher activity completion rates. ICE plots reveal heterogeneous individual responses to engagement features, while chaos-theoretic metrics demonstrate enhanced behavioral stability, reduced sensitivity to initial conditions, and the emergence of persistent usage patterns. Recurrence Quantification Analysis (RQA) further supports the development of structured, repeating behaviors over time. Overall, the findings suggest that the system improvement not only boosted overall engagement but also contributed to a more stable, coherent, and sustainable digital learning environment. This study offers a novel, integrative framework for evaluating e-learning systems by combining explainable AI with nonlinear dynamical systems theory.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Understanding Student Behavioral Dynamics in Moodle Through Chaos Theory and Interpretable Machine Learning

  • 期刊论文

  • 发表日期：

    2025年12月9日

查看全文
  • Mohamed Zine,
  • Fouzi Harrou,
  • Ying Sun

摘要

As digital learning environments continue to reshape higher education, understanding the stability and dynamics of student engagement is essential for optimizing platform design and instructional outcomes. This study examines the behavioral evolution of students utilizing a Moodle-based e-learning system and evaluates the effect of targeted system improvements on platform engagement and learning outcomes. To achieve this, we employ a hybrid analytical framework that integrates interpretable machine learning, namely Individual Conditional Expectation (ICE) and Partial Dependence Plots (PDP), with chaos-theoretic indicators, including Lyapunov Exponent, Fractal Dimension, Butterfly Effect, and Hurst Exponent. Using data collected from 500 university students over a 28-week period, we examine the behavioral changes that followed the introduction of a system improvement introduced at Week 11. The results reveal a substantial shift in engagement patterns following the intervention, including increased login frequency, extended session duration, and higher activity completion rates. ICE plots reveal heterogeneous individual responses to engagement features, while chaos-theoretic metrics demonstrate enhanced behavioral stability, reduced sensitivity to initial conditions, and the emergence of persistent usage patterns. Recurrence Quantification Analysis (RQA) further supports the development of structured, repeating behaviors over time. Overall, the findings suggest that the system improvement not only boosted overall engagement but also contributed to a more stable, coherent, and sustainable digital learning environment. This study offers a novel, integrative framework for evaluating e-learning systems by combining explainable AI with nonlinear dynamical systems theory.