A Machine Learning Framework for Gen-Z College Admission Prediction with a Five-Year Forecast

Abstract

College admission decisions increasingly draw on diverse aca-
demic, extracurricular, and digital-engagement signals, making them a
natural target for data-driven prediction. This study develops a super-
vised machine-learning framework to predict the admission status of
Generation-Z applicants from a large tabular dataset and to forecast
admission-rate trends through 2030. Sixteen raw applicant attributes are
augmented with four engineered composite indicesacademic, extracur-
ricular, digital-readiness, and holisticand three classiers (Random
Forest, Gradient Boosting, and Logistic Regression) are trained and com-
pared using accuracy, the area under the receiver-operating-characteristic
curve (ROC-AUC), and ve-fold cross-validated AUC. On an illustra-
tive run the three models achieve comparable discrimination (AUC ≈
0.760.77), with academic composite, GPA, and standardised-test scores
emerging as the dominant predictors of admission. A quadratic trend
model projects overall, STEM-track, and AI-assisted-screening admis-
sion rates upward over the 20262030 horizon, saturating against an
upper bound. The framework oers an interpretable, reproducible basis
for admission analytics and capacity planning.