An Explainable Model for Identifying At-Risk Student at Higher Education

Abstract

Nowadays, researchers from various fields have shown great interest in improving the quality of learning in educational institutes in order to improve student achievement and learning outcomes. The main objective of this study was to predict the at-risk student of failing the preparatory year at an early stage. This study applies several educational data mining algorithms including RF, ANN, and SVM to build three classification models to meet the objectives of this study. Moreover, different features selection methods namely RFE, and GA have been examined to find the best subset of the highest influential features. Furthermore, several sampling approaches are applied to balance the dataset used in this study, including SMOTE, and SMOTE-Tomek Link. Three datasets related to the preparatory year student from the humanities track at IAU were used in this study. The collected datasets are imbalanced datasets, SMOTE-Tomek Link technique has been used to balance the three proposed datasets. The results showed that RF outperformed other techniques as it records the highest performance for building the models. Moreover, RFE with Mutual Information finds the best subset of features to build the first model. Finally, this study not only developed several classification models to identify at-risk students, but it also went a step further by employing XAI techniques such as LIME, SHAP, and the global surrogate model to explain the proposed prediction models, explaining the output and highlighting the reasons for the student failure.

Advancements in Machine Learning Techniques for Educational Data Mining: An Overview of Perspectives and Trends

“… Increase model generalizability: There is a need to apply the models made by the researchers across many higher education institutes (HEI) to ensure the model is accurate. Researchers are looking into applying a profit-driven approach to the model and use it on other HEIs (Alwarthan et al, 2022;Maldonado et al, 2021). Other researchers wish to apply their model to different courses (Alwarthan et al, 2022).…”

“…Researchers are looking into applying a profit-driven approach to the model and use it on other HEIs (Alwarthan et al, 2022;Maldonado et al, 2021). Other researchers wish to apply their model to different courses (Alwarthan et al, 2022).…”

Physical frailty identification using machine learning to explore the 5-item FRAIL scale, Cardiovascular Health Study index, and Study of Osteoporotic Fractures index

“…SHAP, proposed by Lundberg and Lee in 2017, ( 36 ) is a framework for a unified interpretation of different ML prediction models ( 37 ). It is a Shapley value based on game theory ( 38 ) that explains the impact of each feature on an ML prediction ( 39 ). It is useful for both single- and full-feature interpretability; therefore, it can be used for the entire dataset to explain the influence of each feature on the prediction ( 39 ).…”

Earliest Possible Global and Local Interpretation of Students’ Performance in Virtual Learning Environment by Leveraging Explainable AI

“…The results revealed that Light Gradient Boosted Machine, Logistic Regression, and Gradient Boosting models showed the best results in terms of accuracy, AUC curve, recall, precision, F1-score, and Kappa and Matthews correlation coefficient. Another study was carried out by Sarah A. et al [63] in which an explainable AI model was developed for the identification of students who are at risk of failure in higher education. The SMOTE-Tomek Link technique was utilized for balancing the three imbalanced datasets.…”