PREDVIĐANjE OCENA STUDENATA KORIŠĆENjEM VEŠTAČKIH NEURONSKIH MREŽA

Abstract

The aim of this research is to develop and test an artificial neural network model for predicting first year students’ grades in Basics of ICT at the Faculty of Education, University of Kragujevac, Jagodina. A number of factors could influence students’ grades, such as gender, study programme students are enrolled in (Class teachers, Preschool teachers, Boarding school teachers), tests, seminar papers, proficiency testing in one application software and class attendance. For the purpose of this research the model based on the Levenberg–Marquardt algorithm with the backpropagation was created and trained on the data of the two generations of students at the Faculty of Education in Jagodina, University of Kragujevac. Assessment of the test data has shown that the model was able to correctly predict 90% of future students’ grades. Implications for the application of the model in practice are discussed as well.