Identifying factors affecting learning levels based on Pisa data using machine learning algorithms

The aim of this study was to identify the student, family, and school-level factors that affect the academic performance of Turkish students in terms of scientific literacy, mathematical literacy, and reading skills using PISA 2015-2018 data. Additionally, machine learning algorithms were used to develop models for predicting student performance, and the variables that significantly affected these models were identified. The study also compared the effects of these variables on student performance with those of countries such as Singapore and China. Machine learning algorithms including Random Forest (RF), XGBoost, Support Vector Regression (SVR), and Bayesian Regularized Neural Networks (BRNN) were used to develop prediction models. It was found that XGBoost had a high performance in predicting scores, especially for scientific and mathematical literacy, with low error rates compared to other algorithms. The contributions of the variables to model performance were analyzed using Shapley values, and the most important variables were identified. Furthermore, predictions were made for future school and student scores, which is significant for strategic planning and forecasting the future state of education systems. The findings suggest that current educational policies need to be re-evaluated and improved. The analysis of education data and identifying student- and school-level factors affecting student performance can contribute to making the education system more effective. This process provides a perspective on developing strategies to improve student performance. This study is important for understanding Turkey's current state in PISA exams and creating future educational policies. Additionally, it identified key variables affecting student success, predicted future student and school performances, and contributed to long-term education planning. In this context, the findings are expected to serve as a basis for more detailed and comprehensive future research studies.