Early Detecting Students at Risk Using Machine Learning Predictive Models

Abstract

Analytics in educational environments has received much attention during the last few years. Maintaining a high retention rate is still a major concern in higher educational institutions. Therefore, this research aims to early detect students at risk using three machine learning predictive models, namely support vector machine (SVM), neural networks (NN), and K nearest-neighbors (KNN), based on a new dataset collected from 800 students through surveys. The criteria used to evaluate the models were accuracy, sensitivity, and specificity. Regarding the accuracy, the SVM model has outperformed the NN and KNN models, where it achieves 86.7%. Concerning the sensitivity, the NN model was more sensitive to detect failure cases than the other two models. Regarding the specificity, it was very high for the three models. It is believed that the results could assist the educators in early detecting students at risk, and therefore, reducing students’ dropouts.