Predicting High School Academic Performance in Bosnia and Herzegovina Using Machine Learning Algorithms

This study employs machine learning techniques to predict the academic success of high school students in Canton Sarajevo, aiming to identify key factors influencing performance and assess the accuracy of various predictive models. The results indicate that the Random Forest algorithm demonstrated significant improvements in precision compared to previous studies, highlighting its potential for use in educational decision-making. Key predictors of student performance included prior academic success, subject-specific grades, parental education, and engagement in academic activities such as homework and class participation. The study’s practical implications suggest that educators and policymakers can use machine learning models to develop targeted interventions, improve student guidance, and implement early warning systems for at-risk students.