Using Machine Learning models to predict the on-time graduation status of students

Abstract

The study aims to perform optimal Machine Learning model selection to predict the on-time
graduation status of students. By using the dataset of students majoring in Banking faculty from the Banking
Academy during the period of 2010-2020 through Machine Learning models such as Logistic Regression,
K-Nearest Neighbors, Decision Tree, Random Forest, Support Vector Machine, XGBoost, and CatBoost, the
study has chosen Random Forest as the optimal model. The research has identified 2 attributes: Academic
processing information and Grade Point Average (GPA) of semesters 1 through 4 have a strong impact on
the ability of students to graduate on time or late, and proposed some recommendations to help the school
provide solutions to improve the graduation rate of students.