Machine Learning for Predicting Students’ English Test Scores in an Educational Setting

Machine learning methods have been increasingly used in educational settings. For the problem of predicting students’ English test scores, first, the number of absences and the average score of usual tests were selected as students’ behavioral features; then, the prediction effects of five machine learning approaches, including K-medoids, support vector regression (SVR), random forest (RF), gradient boosting decision tree (GBDT) and XGBoost, were compared after eliminating the features with weak correlation. The results showed that the XGBoost method best predicted students’ English test scores, with an average accuracy, precision, and recall rate of 0.8619, 0.8322, and 0.8734, respectively, and an F1 value of 0.8523. The findings demonstrate the reliability of the XGBoost method in predicting students’ English test scores. It can be extended and applied in practice. This article makes some contributions to the accurate prediction of students’ grades and the intelligent management of educational information.