Integrated Computational Learning Algorithm for Undergraduates’ Academic Early Warning

Abstract

With the development of education big data, it is helpful for education managers to use machine learning method to predict students’ academic warning status, in order to ensure that students pass the course and graduate on time. Although significant progress on predicting academic warning status has been achieved in recent years, existing methods are often lack of generalization and are difficult to be applied in real scenarios. In this study, we divide students’ scores into five coarse-grained dimensions: mathematics, foreign language, humanities, major and total score, and innovatively use machine learning ensemble model to predict college students’ academic status. Experiments show that the coarse-grained dimension division is not only conducive to the generalization of the model, but also improves the accuracy of prediction academic warning status by 9.708%. Meanwhile, experimental results show that the multi machine learning model ensemble technique can effectively improve the prediction accuracy of college students’ academic warning status. With only a small number of samples, the accuracy of predicting students’ academic warning status by three months in advance reaches 97.52%, and that by the previous semester reaches above 97.93%.