Machine Learning Techniques for Predictive Analytics of Academic Outcomes and Behavior of Students

Abstract

This study looks at how machine learning (ML) can be used to look at and guess how well secondary school pupils in two Portuguese schools will do in school. The study uses a dataset with 33 variables that include a lot of demographic, social, and academic elements, such as age, gender, study time, family history, and the education levels of the parents. To make it easier to sort the data, the objective variable students' final grade (G3) was changed into a binary outcome: students who scored 10 or higher were put in the "pass" (1) group, and those who scored less than 10 were put in the "fail" (0) group. We used three machine learning models which are: Random Forest, Logistic Regression, and Gradient Boosting to see how well they could predict. Random Forest had the best accuracy of the three models, at 92.4%, followed by Logistic Regression at 89.9%, and Gradient Boosting at 86.08%. The study shows that past academic performance and study habits are important indicators of how well a student will do. The results show that machine learning could be very useful in schools, which is helpful information for teachers, school administrators, and politicians. These ideas can help build focused interventions, personalized learning plans, and early warning systems that will help students do better and lower the number of students who fail in secondary school.