Neural Networks for predicting student performance in online education

Failure rates in online higher education raises as a major problem for the modal instruction, thus, prediction of student performance is proposed as a preventive strategy to diminish student failure. Research in student performance prediction is of a wide variety that complicates the replication of studies in order to take advantage of research results. This study tackles this issue by proposing three types of neural networks for student performance prediction, which are built from standardized variables more easily obtained than those used in most of studies identified. By the statistical comparison of the prediction accuracies of the neural networks with that of a statistical linear regression, the three neural networks obtained higher prediction accuracy. As conclusion, neural networks are proposed as techniques for an early prediction and identification of students in risk of failure.