Supporting Student Success Through an Interpretable Early Warning System in a First-Year Statistics Unit

Abstract

Early identification of students experiencing academic difficulty is critical to supporting student success and retention in their first year of university. This study examines an interpretable learning analytics framework designed to support timely intervention in a first-year introductory statistics unit. Data from 566 students across three cohorts were used to develop and evaluate predictive models using assessment results collected during the teaching period and enrolment characteristics. The selected model demonstrated strong predictive performance and was validated on a subsequent cohort. To support practical implementation, predictions were translated into a colour-coded risk report that enables educators to quickly identify and prioritise students requiring academic support. SHapley Additive exPlanations (SHAP)-based explanations were incorporated to provide transparency and assist educators in understanding the factors contributing to individual risk classifications. Findings indicate that formative assessments are the strongest indicators of academic success, highlighting opportunities for proactive and targeted intervention. The framework demonstrates how interpretable learning analytics can support evidence-informed decision-making and student success in large first-year units.