Prediction of Primary and Secondary Education Institutions Scholarship Examination (PSEISE) Success with Artificial Neural Networks
Prediction of Primary and Secondary Education Institutions Scholarship Examination (PSEISE) Success with Artificial Neural Networks

Summary/Abstract: Background/purpose. This study aims to reveal in detail the extent to which the variables in The Primary and Secondary Education Institutions Scholarship Examination (PSEISE) predict the success of students on the scholarship exam with the help of artificial neural networks (ANN). In addition, in light of the findings obtained as a result of the research, it aims to contribute to improving the quality and content of PSEISE and to revising the variables in student selection. Materials/methods. A descriptive-relational screening model and purposive sampling method were used in the study. The study group of the research included students who were studying in a provincial centre in the Aegean Region of Turkey in the 2023-2024 academic year and who took the PSEISE at the 5th-grade level. In the study, the prediction level of these students’ 4th and 5th grade written exam scores (independent variable) of Turkish, Mathematics, Science, Social Sciences, Religious Culture, and Ethics courses (independent variable) and their PSEISE achievement status (dependent variable) were determined through MATLAB 2023a program classification interface and optimised to achieve the best result. Results. This process, carried out with a 5-fold cross-validation method, concluded that the ANN model with sigmoid activation function, forward cascaded back propagation, and double layer (230 neurons in the first layer and 2 neurons in the second layer) has the highest performance and an accuracy rate of 91.7%. Conclusion. The correct classification of unsuccessful students with high performance shows that the model effectively predicts these students. However, the lower accuracy rates obtained in predicting successful students indicate that additional variables (taking private lessons and courses, etc.) should be examined to improve the performance of the ANN model for this group.

• Details
• Contents