Artificial neural networks in academic performance prediction: Systematic implementation and predictor evaluation

Abstract

Abstract The applications of artificial intelligence in education have increased in recent years. However, further conceptual and methodological understanding is needed to advance the systematic implementation of these approaches. The first objective of this study is to test a systematic procedure for implementing artificial neural networks to predict academic performance in higher education. The second objective is to analyze the importance of several well-known predictors of academic performance in higher education. The sample included 162,030 students of both genders from private and public universities in Colombia. The findings suggest that it is possible to systematically implement artificial neural networks to classify students’ academic performance as either high (accuracy of 82%) or low (accuracy of 71%). Artificial neural networks outperform other machine-learning algorithms in evaluation metrics such as the recall and the F1 score. Furthermore, it is found that prior academic achievement, socioeconomic conditions, and high school characteristics are important predictors of students’ academic performance in higher education. Finally, this study discusses recommendations for implementing artificial neural networks and several considerations for the analysis of academic performance in higher education.