Nourishing the Body and Mind of University Students: Using a Machine Learning Approach to Prioritize Outreach Strategies for a Campus Food Pantry

Food insecurity (FI) may lead to lower academic achievement, yet college students with inadequate food underutilize campus food pantries. This research aimed to identify predictors of academic success among pantry shoppers (PSs) to inform outreach. Data from AY 2021–2022 (N = 847) and 2022–2023 (N = 951) were derived from swipes of student identification cards, merged with university student-provided data, and de-identified. Multiple regression, logistic regression, and Least Absolute Shrinkage and Selection Operator (LASSO) were employed to create and validate models using Machine Learning. Grade Point Averages (GPAs) were compared by two-sample t tests. The PSs demonstrated higher GPAs in the fall term than non-pantry shoppers (p = 0.04). Validation of the models indicated strong performance. Multiple regression yielded a low prediction error (0.05), and logistic regression achieved 71% accuracy (AUC = 0.776). LASSO identified positive predictors of academic success, including graduate and honors status, junior and senior classification, females, international residency, and frequency of pantry shopping. Negative predictors included part-time status, first-year status, Black or Hispanic ethnicity, and Pell Grant eligibility. Findings underscore the complex interplay between sociodemographic and academic factors that should be considered when planning pantry outreach programs and highlight the need for standardized measures of student pantry utilization, which may aid resource allocation and sustainability.