Advancing educational data mining for enhanced student performance prediction: a fusion of feature selection...

Area of Science:

• Educational Data Mining (EDM)
• Machine Learning in Education
• Learning Analytics

Background:

• Educational institutions generate vast amounts of data, presenting opportunities for EDM to enhance learning outcomes.
• Traditional feature selection methods often rely on static thresholds, which can be insufficient for evolving educational datasets.
• Overfitting and underfitting are common challenges in predictive modeling within educational contexts.

Purpose of the Study:

• To introduce a novel feature selection model, Dynamic Feature Ensemble Evolution for Enhanced Feature Selection (DE-FS).
• To address limitations of static feature selection methods through dynamic and adaptive thresholding.
• To improve the accuracy and flexibility of predicting student performance.

Main Methods:

• DE-FS combines traditional methods (correlation matrix analysis, information gain, Chi-square) with heat maps for feature selection.
• A core innovation is a dynamic and adaptive thresholding mechanism that adjusts based on evolving data patterns.
• The model's predictive performance was evaluated across diverse educational datasets.

Main Results:

• DE-FS demonstrated superior predictive performance compared to traditional methods.
• The dynamic thresholding mechanism effectively adapted to fluctuating data patterns.
• The model achieved precise and reliable predictions of student performance.

Conclusions:

• DE-FS offers an advanced ensemble-based feature selection methodology for educational data mining.
• The adaptive thresholding enhances model accuracy, flexibility, and robustness.
• DE-FS supports targeted interventions and improved resource allocation for personalized learning experiences.