Role of convolutional features and machine learning for predicting student academic performance from MOODLE data.

Area of Science:

• Educational Data Mining
• Artificial Intelligence
• Machine Learning

Background:

• Predicting student performance is crucial due to large educational datasets.
• Existing methods in educational data mining (EDM) struggle with accuracy, imbalanced data, and feature engineering.
• Learning platforms can analyze student data to improve outcomes and reduce failure rates.

Purpose of the Study:

• To propose a machine learning system for accurate student academic performance prediction.
• To address challenges in existing EDM techniques, including data imbalance and feature engineering.
• To evaluate the effectiveness of deep convoluted features against original features.

Main Methods:

• Developed a machine learning-based system utilizing deep convoluted features.
• Employed the synthetic minority oversampling technique (SMOTE) for handling imbalanced datasets.
• Evaluated performance using both original and deep convoluted features with an extra tree classifier.

Main Results:

• Deep convoluted features significantly improved prediction accuracy compared to original features.
• The extra tree classifier with convoluted features achieved a classification accuracy of 99.9%.
• The proposed AI-driven system outperformed state-of-the-art approaches in student performance prediction.

Conclusions:

• The proposed system offers substantial advancements in AI-driven student performance prediction.
• Deep convoluted features are highly effective for improving accuracy in educational data mining.
• This research provides a powerful tool for identifying at-risk students and enhancing learning.