Abstract:

Reliable evaluation of predictive algorithms in Learning Analytics is often constrained by limited access to real learner data, inconsistent logging practices across courses, and strict ethical requirements that restrict data sharing. SimLearn introduces a technical sandbox framework that integrates directly with Moodle to generate synthetic yet structurally authentic learner activity data. The system enables controlled experimentation by emulating student behaviour with high configurability and reproducibility.

The sandbox operates through a simulation engine capable of modelling behavioural profiles based on parameters such as engagement intensity, temporal study patterns, task completion strategies, error rates, procrastination tendencies, and dropout likelihood. These profiles drive automated interactions with Moodle components including assignments, quizzes, forums, SCORM packages, and resource views. By leveraging Moodle’s native logging architecture, SimLearn produces event streams identical in format to real log data (e.g., logstore_standard_log entries), ensuring seamless compatibility with existing Learning Analytics pipelines.

This technical design allows researchers to execute structured experiments to benchmark and stress-test predictive models, such as classification algorithms for early dropout detection, regression models for performance forecasting, and clustering methods for identifying behavioural archetypes. SimLearn further supports scenario-based evaluations, enabling users to vary course structures, workload distributions, and assessment types to analyse how algorithmic performance changes under different pedagogical conditions. The framework also incorporates mechanisms to inject controlled noise, simulate missing data, and manipulate event density to assess the robustness and generalisability of predictive approaches. Because all generated data is synthetic, the sandbox eliminates privacy concerns and facilitates cross-institutional collaboration, algorithm comparison, and reproducibility in Learning Analytics research.

Beyond methodological testing, SimLearn serves as a training environment for educators, allowing them to explore LA dashboards, interpret behavioural patterns, and understand the dynamics of algorithmic predictions without exposing real student data. As a modular and extensible simulation platform, SimLearn contributes to advancing transparent, ethical, and technically rigorous Learning Analytics in Moodle-based learning ecosystems.

Keywords:

Learning Analytics, Synthetic Learner Data, Behaviour Simulation, Moodle Sandbox, Predictive Modelling, Early-Warning Systems, Educational Data Mining.