Research / Learning analytics at scale

Which course choices set students up to succeed?

At Monash University, I analyzed more than 7 million course enrollment records to give faculty evidence for curriculum decisions, and built tools to support students at risk of falling behind.

Learning analytics at scale, 2020 to 2023Curriculum analyticsPredictive modelingDashboardsPython and R

Dot plot comparing high-performing and low-performing students. For each course, it shows when in their degree students took it, how many took it, and how difficult it was for that group. — When high- and low-performing students took each course, and how difficult it was for them.

The problem

Universities make curriculum decisions, such as which subjects to require, recommend, or sequence, with little evidence about how those choices affect students. At the same time, students who are falling behind are often identified too late to help.

What I did

Analyzed historical records of every course in the university, more than 7 million student enrollments, to study how subject choices relate to academic performance.
Built early-warning models (random forests and logistic regression) to identify at-risk students, and designed feedback interventions to support them.
Designed a curriculum analytics dashboard in R Shiny so faculty could explore program-level patterns themselves.

What I found

The courses students chose, and when they took them, were linked to their academic performance, which pointed to specific pathways and bottleneck courses worth redesigning.

Why it matters

The Faculty of Information Technology adopted the recommendations from this work to improve its teaching practices. It showed how institutional data, used carefully, can move curriculum decisions from intuition to evidence.

Publications

Curriculum analytics of course choices: Links with academic performance. Journal of Learning Analytics, 2024