OBE-Based Course Outcomes Prediction Using Machine Learning Algorithms

Educational data mining has emerged as a powerful tool for exploring hidden patterns in student data, predicting academic success, and reducing the dropout rate, especially in higher education. The OBE approach has been widely used in almost all educational institutions to ensure that students have the competencies determined by the study program and to measure student achievement. This study proposes a prediction model using machine learning algorithms using Regression to predict student course outcomes in an integrated OBE attainment system utilizing midterm component scores. This study identifies the most potential machine learning techniques for predicting learning outcomes in an OBE-based framework. By predicting student course outcomes early, it is possible to identify at-risk students who fail to meet desired outcomes early so that lecturers can take preventive actions. The performance of the decision trees, random forests, support vector, and K-nearest neighbor algorithms was computed and compared to predict the students' course outcome results. This study uses a dataset of 2423 students enrolled in 30 courses in 6 study programs at a private university in Indonesia during the first semester of 2023–2024. The predictions were made using mid-term course components' scores defined in the course learning plan, such as assignments, class engagement, quizzes/exams, and case-based projects. This research finds support vector regression recommended for use in large classes, while decision trees and random forests are more suitable for smaller classes.