Identifying At-Risk Students

Identifying At-Risk Students Using Predictive Analytics

The Learning Analytics team are running two collaborative projects in 2021 supporting staff in identifying at-risk students using predictive analytics and risk factors. Following a successful pilot last year, we are seeking course coordinators interested in identifying at-risk students within their course(s) in 2021. In particular, one project involves using early (pre-census) assessment results to predict final course outcomes (pass/fail). This project generally requires courses that have:

  • Consistent assignment structure across multiple deliveries of the course
  • Assessment results recorded in MyUni
  • Early assessment (ideally pre-census)

If your course does not meet the criteria above, then it may instead be suited to our at-risk dashboard. This dashboard assigns risk statuses to students within a course/program. We will work collaboratively with individual course coordinators to select risk flags based on the course design. Examples of default risk flags include student learning engagement (e.g. overdue assessments; course access frequency; use of learning resources) as well as other study outcomes (e.g. low GPA; previous course attempts).

The insights provided in both of these projects can help teaching staff target messaging to at-risk students, which has been found to lead to increased engagement and performance within the course. These insights will also be used by student support services like Succeed @ Adelaide to inform the UoA strategies for engaging and supporting students.

 

How do I get involved?

Please send an email to the Learning Analytics team. We can assess your course to see which project is best suited.

I want to know more about the 2020 Pilot

The pilot involved running the predictive model in several courses in Semester 2, 2020. Our model predicted final course outcomes (pass/fail) for each student within the course, based on early assessments within those courses. The final accuracy for the predictive model ranged from 87% to 96% in predicting student course outcomes across the various courses.

Based on these predictions, some course coordinators opted to contact at-risk students to offer support and discuss any issues they were having. These interactions lead to improved engagement amongst these students, and in many cases better performance and a passing final grade.