Using neural networks to predict the future performance of students

Abstract

This work studies two models of artificial neural networks (ANN) in order to predict at the beginning of the school year the performance of students in SPAECE (Permanent System of Evaluation of Basic Education of Ceará). The factors considered as relevant to student achievements were the grade, period of study, school score and student score in elementary school, which were used as input variables to an ANN. A model of Perceptron and Multilayer Perceptron (MLP) were developed and trained using data of students of the year 2013 from schools of the 6th CREDE (State Development Regional Coordination). Test data evaluation shows that the ANN model is able to correctly predict 95% of performance of beginning students.