Learning Analytics Project: Student Engagement & Academic Outcomes

Python | Tableau | Statistical Analysis | Longitudinal Analysis | A/B Test Design

Analyzed 32,593 student-course records and 10.6M+ LMS interactions to examine student engagement, academic performance, and opportunities for early learner-support interventions.

Research Question

How does online learning engagement relate to academic outcomes, and when do disengagement patterns emerge?

Business Purpose

Identify engagement patterns to inform early interventions, learner retention, and data-driven learning strategies.

Methodology

• Data Preparation: Analyzed student records, LMS interactions, assessment scores, and course outcomes using Python.
• Statistical Analysis: Applied descriptive statistics, one-way ANOVA, and Pearson correlation to examine engagement and performance.
• Longitudinal Analysis: Compared weekly engagement patterns across Distinction, Pass, Fail, and Withdrawn students.
• Experimental Design: Proposed an A/B test to evaluate proactive learner-support messages.

Findings

• Engagement Differences: Distinction students averaged 2,667 LMS clicks versus 445 for Withdrawn students; differences across outcome groups were statistically significant (p < .001).
• Academic Performance: Engagement was positively correlated with assessment scores (r = 0.293, p < .001).
• Engagement Trends: Students who failed or withdrew generally showed lower weekly LMS engagement than those who passed or earned distinctions.

Key Insights and Recommendations

• Early Intervention: Identify students with low LMS engagement for targeted learning support.
• Personalized Support: Pilot reminders, study-planning prompts, and relevant course resources.
• Impact Evaluation: Use randomized A/B testing to assess subsequent engagement and assessment completion before scaling interventions.

Data Source

Open University Learning Analytics Dataset (OULAD)

Real-world, anonymized educational data covering 7 courses, 22 course presentations, 32,593 student-course records, and 10.6M+ LMS interactions.

Code

• Python (Pandas, SciPy, Matplotlib): Data preparation, descriptive statistics, ANOVA, correlation analysis, and longitudinal engagement analysis.
• View Python Analysis — Jupyter Notebook

Data Visualizations and Dashboard

• Tableau: Interactive dashboard comparing LMS engagement, academic performance, and weekly engagement trends across course outcomes.
• View Interactive Tableau Dashboard

Limitations

• Observational findings establish associations, not causation.
• LMS clicks measure activity, not learning quality.
• The A/B test is proposed, not implemented.