Educational data mining has become an effective tool for exploring the hidden relationships in educational data and predicting students' academic performance. Educational evaluation is an important part of the teaching process, and the traditional evaluation methods have problems such as high subjectivity and low efficiency. The advancement of machine learning technology has led to an increasing interest in data-driven methods for evaluating student teaching. This study utilizes a dataset pertaining to student evaluations from Turkey to apply and compare ten machine learning algorithms, namely Random Forest (RF), Gradient Boosting Machine (GBM), Naive Bayes (NB), K-Nearest Neighbors (KNN), Neural Networks Algorithm (Nnet), Flexible Discriminant Analysis (FDA), Support Vector Machine (SVM), Classification and Regression Trees (CART), Sparse Linear Discriminant Analysis (SLDA), and AdaBoost (ADA), in predicting student satisfaction ratings. The findings indicate that the SVM algorithm yielded the most favorable results, with performance metrics including Accuracy, Sensitivity, Specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), Precision, Recall, F1 Score and Standard Error (SE) recorded at 0.9765, 0.9887, 0.9789, 0.9891, 0.9789, 0.9765, 0.9777, and 0.0042, respectively. Furthermore, the SVM model's predictive outcomes were elucidated by applying the SHAP (SHapley Additive exPlanations) framework. Finally, we developed the Shiny application for online prediction of learning effect satisfaction, which can provide educators with a scientific basis for evaluation and technical support for personalized teaching and curriculum optimization.
Zhan et al. (2025) studied this question.