Development of a prediction model for student teaching satisfaction based on 10 machine learning algorithms.

Area of Science:

• Educational Data Mining
• Machine Learning in Education

Background:

• Traditional educational evaluation methods suffer from subjectivity and inefficiency.
• Advancements in machine learning offer data-driven solutions for student teaching evaluation.
• Educational data mining reveals hidden patterns and aids in predicting student performance.

Purpose of the Study:

• To compare ten machine learning algorithms for predicting student satisfaction ratings.
• To identify the most effective algorithm for evaluating student teaching.
• To develop a practical tool for educators based on predictive modeling.

Main Methods:

• Applied ten machine learning algorithms: Random Forest, Gradient Boosting Machine, Naive Bayes, K-Nearest Neighbors, Neural Networks, Flexible Discriminant Analysis, Support Vector Machine (SVM), Classification and Regression Trees, Sparse Linear Discriminant Analysis, and AdaBoost.
• Utilized a dataset of student evaluations from Turkey.
• Employed the SHAP (SHapley Additive exPlanations) framework to interpret SVM model predictions.
• Developed a Shiny application for online prediction of learning effect satisfaction.

Main Results:

• The Support Vector Machine (SVM) algorithm demonstrated superior performance in predicting student satisfaction.
• SVM achieved high accuracy (0.9765) and other key performance metrics like Sensitivity (0.9887) and Specificity (0.9789).
• The SHAP framework provided insights into the SVM model's predictive drivers.

Conclusions:

• Machine learning, particularly SVM, offers a robust and accurate approach to educational evaluation.
• The developed Shiny application provides educators with a scientific basis for evaluation and personalized teaching support.
• Data-driven insights can optimize curriculum and enhance the learning experience.