Combination prediction method of students' performance based on ant colony algorithm.

Area of Science:

• Educational Technology
• Artificial Intelligence
• Machine Learning

Background:

• Student performance is crucial for evaluating teaching quality and guiding learning.
• Single prediction models often lack sufficient accuracy for student performance analysis.
• Existing methods may not efficiently integrate diverse predictive capabilities.

Purpose of the Study:

• To develop a superior student performance prediction model.
• To address the accuracy limitations of individual machine learning models.
• To leverage ant colony optimization for enhanced predictive performance.

Main Methods:

• Selected Decision Tree (DT), Support Vector Regression (SVR), and BP Neural Network (BP) for individual models.
• Employed Ant Colony Algorithm (ACO) to determine optimal weights for model combination.
• Evaluated the combined model against single models and other advanced methods.

Main Results:

• The ACO-based combination model achieved a Mean Square Error (MSE) of 0.0089, significantly outperforming DT (0.0326), SVR (0.0229), and BP (0.0148).
• The combined model demonstrated superior performance compared to GS-XGBoost (MSE 0.0131), PSO-SVR (MSE 0.0117), and IDA-SVR (MSE 0.0092).
• The proposed method exhibited a faster running time than the compared advanced prediction models.

Conclusions:

• The ant colony algorithm-based combination prediction model offers a significant improvement in student performance prediction accuracy.
• This hybrid approach effectively integrates multiple machine learning models for robust educational data analysis.
• The method provides a computationally efficient and accurate tool for timely intervention and support in student learning.