Classification of mathematical test questions using machine learning on datasets of learning management system...

Area of Science:

• Educational Technology
• Machine Learning in Education
• Artificial Intelligence in Learning Management Systems

Background:

• Students exhibit varied mathematical proficiencies, necessitating tailored educational content.
• Learning Management Systems (LMS) leverage AI and databases to personalize online learning experiences.
• Classifying question difficulty is crucial for LMS to recommend appropriate content and boost learning efficiency.

Purpose of the Study:

• To classify large-scale mathematical test items by difficulty level using machine learning techniques.
• To identify significant variables correlating with question difficulty.
• To evaluate the performance of different machine learning models for question difficulty classification.

Main Methods:

• T-test analysis to identify significant correlations between variables (answer rate, question type, solution time) and difficulty level.
• Classification of mathematical items using machine learning models: Logistic Regression (LR), Random Forest (RF), and XGBoost.
• Evaluation of models using metrics such as accuracy, precision, recall, F1 score, AUC-ROC, Cohen's Kappa, and MCC.

Main Results:

• Answer rate, question type, and solution time were found to be significantly correlated with question difficulty.
• The XGBoost model demonstrated superior performance, achieving 85.7% accuracy and 85.8% F1 score.
• Machine learning models effectively classified mathematical test items based on difficulty.

Conclusions:

• Machine learning, particularly XGBoost, provides an effective method for classifying mathematical question difficulty.
• These classifications can serve as an auxiliary tool for LMS to recommend suitable questions to learners.
• Personalized question recommendations based on difficulty can enhance overall learning efficiency.