Student Graduation Prediction Using Machine Learning Algorithms: Application of Linear and Logistic Regression on Educational Factors

Abstract

The main challenge in education lies in accurately predicting student graduation and understanding the factors influencing it. This study aims to provide a data-driven solution using machine learning algorithms, specifically linear regression to predict student exam scores and logistic regression to classify student graduation. The study contributes by developing predictive models that serve as tools to support strategic decision-making in educational institutions. This study utilized the Student Performance Factors dataset, comprising 6,607 samples with independent variables such as study hours, attendance, and parental involvement. Data analysis involved cleaning, transformation, and normalization before applying regression models. The findings showed that linear regression achieved a Mean Squared Error (MSE) of 3.256, indicating high accuracy in predicting exam scores. Logistic regression demonstrated an accuracy of 99.85% in classifying student graduation. These models complement each other by offering strategic insights to enhance educational quality.