Regularized multi-path XSENet ensembler for enhanced student performance prediction in higher education

Area of Science:

• Educational Data Mining
• Machine Learning
• Artificial Intelligence

Background:

• Increasing educational data necessitates advanced predictive models for institutional planning and student support.
• Forecasting student performance (low, medium, high) is crucial for timely interventions.

Purpose of the Study:

• To develop and evaluate XSEJNet, a novel hybrid model for accurate student performance prediction.
• To enhance predictive accuracy and computational efficiency in educational data mining.

Main Methods:

• Proposed XSEJNet: a hybrid model integrating ResNeXt architecture with squeeze-and-excitation (SE) attention mechanisms.
• Utilized Jaya optimization algorithm for hyperparameter tuning.
• Analyzed structured and unstructured academic data to capture high-dimensional features.

Main Results:

• XSEJNet achieved a prediction accuracy of 97.98%.
• Outperformed conventional machine learning and advanced techniques like RLCHI, AEO-XGBoost, Conv-DL, and DualGNN.
• Demonstrated faster convergence and reduced computational overhead compared to existing methods.

Conclusions:

• XSEJNet offers a scalable and practical solution for real-world educational settings.
• Supports early intervention, enhances e-learning platforms, and informs institutional decision-making.
• Contributes to developing inclusive, data-driven, and sustainable academic systems.