2026 ASEE Annual Conference & Exposition

Attainment of complex student outcomes, particularly teamwork and experimentation and data analysis, is essential for Accreditation Board for Engineering and Technology (ABET) accreditation and for preparing civil engineering graduates for professional practice. However, many ABET assessment systems rely on lagging indicators derived from end-of-term evaluations, limiting the ability of programs to implement timely instructional and advising interventions under ABET Criterion 4 Continuous Improvement. This study investigates the use of outcome-level predictive analytics to forecast end-of-term attainment of complex civil engineering student outcomes using early, course-embedded assessment data.
Using a multi-year dataset spanning Fall 2021 through Spring 2025, comprising over 31,000 course-embedded ABET outcome assessment records aggregated to the student-term level, supervised machine learning models were developed to predict attainment of teamwork (E5) and experimentation and data analysis (E6) outcomes. Predictor features were derived exclusively from early outcome-level assessment submissions available during the first 30 percent of each academic term. Logistic Regression and Random Forest classifiers were evaluated using five-fold stratified cross-validation, with performance assessed through accuracy, Area Under the Receiver Operating Characteristic Curve (AUC), and class-specific F1-scores.
Results demonstrate strong and consistent predictive performance across outcomes. For experimentation and data analysis, models achieved AUC values near 0.90, indicating robust discrimination under class imbalance. For teamwork outcomes, AUC values approached 0.99, reflecting near-perfect separation under the present data conditions. Across both outcomes, Random Forest models provided modest gains in accuracy, while Logistic Regression offered competitive discrimination with greater interpretability. Exploratory fairness evaluations were incorporated to support responsible and equitable application of predictive analytics.
The findings demonstrate that early, outcome-level assessment data contain meaningful predictive signal and can be leveraged to strengthen ABET Criterion 4 continuous improvement processes. By relying exclusively on data elements already collected by most ABET-accredited programs, the proposed predictive analytics framework is scalable, transparent, and adaptable across civil engineering curricula, enabling earlier intervention and more effective, equity-aware student support.