Multi-Layer Perception model with Elastic Grey Wolf Optimization to predict student achievement.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Educational Data Mining

Background:

• Student achievement prediction is crucial for educational improvement.
• Existing prediction models may lack accuracy and stability.
• Optimization algorithms are key to enhancing predictive model performance.

Purpose of the Study:

• To introduce a novel optimization algorithm, Elastic Grey Wolf Optimization (EGWO).
• To develop an EGWO-optimized Multi-Layer Perception (EGWO-MLP) model for student achievement prediction.
• To evaluate the effectiveness of the EGWO-MLP model using real-world student data.

Main Methods:

• Development of the Elastic Grey Wolf Optimization (EGWO) algorithm with specific mechanisms.
• Application of EGWO to optimize weights and biases of a Multi-Layer Perception (MLP) model.
• Training and validation of the EGWO-MLP model on the UCI student performance dataset.

Main Results:

• The EGWO-MLP model demonstrated superior prediction accuracy for Mathematics achievement compared to baseline models.
• For Portuguese achievement, the EGWO-MLP model outperformed multiple models in both training and testing phases.
• The EGWO-MLP model exhibited reduced test errors, indicating effective weight/bias feedback and strong exploration capabilities.

Conclusions:

• The proposed EGWO algorithm effectively optimizes MLP models for student achievement prediction.
• The EGWO-MLP model is a feasible and robust tool for predicting student academic success.
• This approach offers valuable insights for enhancing educational strategies and teaching quality.