Evaluating Machine Learning Algorithms for Predicting Financial Aid Eligibility: A Comparative Study of Random Forest, Gradient Boosting and Neural Network

Abstract

Financial aid ensures equitable access to higher education, irrespective of students' social or economic backgrounds. However, as the financial aid source are limited, the administrators are burdened with the task of determining the student eligibility for financial aid in a fair and unbias manner. Additionally, the process of determining eligibility by human evaluators can benefit from machine learning assisted decision support tools. This study investigates the feasibility of using machine learning algorithms to achieve this goal. Three algorithms were selected for this comparative study with Decision Tree acts as a baseline. The algorithms are trained against a highly imbalanced dataset provided by ZAWAF. The training process incorporates k-fold cross-validation and employs stratified sampling techniques. It was found that all machine learning models outperformed the baseline, with the MLP-ANN model exhibiting the highest accuracy and precision scores. This demonstrates the potential for machine learning models to be integrated as decision support for the distribution of financial aid.