What factors enhance students' achievement? A machine learning and interpretable methods approach.

Area of Science:

• Educational Technology
• Learning Analytics
• Artificial Intelligence in Education

Background:

• Traditional research often overlooks the complex interactions influencing student achievement.
• Existing studies frequently analyze isolated factors or simple correlations, missing multivariate relationships.

Purpose of the Study:

• To model the multivariate relationships between behavioral and instructional predictors and student achievement.
• To utilize interpretable AI to uncover nuanced factor-achievement dynamics.
• To provide actionable insights for improving pedagogical strategies and student support.

Main Methods:

• Employed an ensemble of five machine learning algorithms (SVM, DT, ANN, RF, XGBoost).
• Modeled relationships between four behavioral and six instructional predictors using final exam performance as the outcome.
• Applied interpretable AI techniques to identify key patterns and factor contributions.

Main Results:

• Machine learning with explainability effectively identified nuanced factor-achievement relationships.
• Behavioral metrics (homework, answering, discussion, attendance scores) consistently showed positive associations with achievement.
• High-achieving students exhibited stronger collaborative skills and a preference for technology-enhanced learning environments.
• Gamification frequency positively impacted outcomes, whereas assignment frequency had counterproductive effects.

Conclusions:

• Educators should balance direct instruction with active learning modalities to optimize student achievement.
• Predictive analytics, leveraging identifiable learning features, can inform early warning systems for proactive student support.
• The developed framework facilitates the transformation of predictive analytics into practical pedagogical improvements.