Analysis of student achievement in high school entrance exam with machine learning techniques

In recent years, Machine Learning algorithms have become widely used in Educational Data Mining studies. This study aims to predict the success in the High School Transition Exam (LGS) using a dataset that includes students' academic performance scores as well as individual and environmental characteristics. The study also proposes the best model by comparing the performances of both single-base and ensemble algorithms. In the comparison of the models, the R-Squared value is considered as a reference, while metrics such as MSE, RMSE, MAE, and MAPE are also examined. Additionally, the study investigates the impact of Feature Selection methods on the model's performance. The results show that the best performance is achieved by utilizing the Super Learner model in a Stacking architecture, using the entire dataset without applying Feature Selection. The model's R-squared value is found to be 0.79, and the MSE is calculated as 161.32. This study demonstrates that the model is open to further improvements based on the obtained results. Moreover, it indicates that the success levels in central exams can be modeled by utilizing data that includes students' educational and personal characteristics. This technique, used in the study, can play an important role in identifying students' strengths and weaknesses and improving their academic performance in the future.