Student success in Science, Technology, Engineering, and Mathematics (STEM) is a national concern. To increase engineering retention and graduation rates at a small private institution, a university council developed a binary classifier to identify high-risk students and proposed interventions that included decoupling first-year Physics and Calculus courses, support in introductory Calculus, and Spatial Visualization (SV) training. This paper aims to validate the binary classifier used to identify the under-prepared students entering their first year and assess the impact of the interventions. We provide a comparative analysis of student success metrics for high-risk engineering students across a decade of cohorts, including 5 years before (2006–2010) and 5 years after (2011–2015) implementation of intentional strategies.
We validated the binary classifier using an accuracy measure and Matthews Correlation Coefficient (MCC). We used the 2-population proportion test to compare STEM retention and 4- and 6-year graduation rates of High-Risk engineering students before and after interventions and compare student performance in early foundation STEM courses across the same time frame.
The binary classification model identified High-Risk students with an accuracy of 63–70% and an MCC of +0.28 to +0.30. In addition, we found statistically significant improvement (
The methodology and strategies presented may provide effective guidance for institutions seeking to improve the overall performance of undergraduate students who otherwise might struggle in their first-year engineering curriculum.