Swarm Optimized S-AAF: Enhancing Student Academic Performance Prediction with Swarm-Optimized Adaptive Binary Classifier

Abstract

Introduction: Predicting academic performance is crucial for educational institutions to implement timely interventions and strategies that enhance student success. Traditional predictive models often struggle with the nonlinear and high-dimensional characteristics of educational data, leading to suboptimal outcomes.

Objectives: This research aims to develop a more effective predictive framework that overcomes the limitations of conventional models by leveraging advanced optimization and classification techniques.

Methods: The work proposes the Swarm Optimized S-AAF model, which integrates Particle Swarm Optimization and Genetic Algorithm with a robust classifier called Sigmoid-plus Adaptive Activation Function . Particle Swarm Optimization optimizes parameters based on swarm intelligence, while GA uses evolutionary strategies to refine the solution space. Together, these algorithms enhance feature selection and improve the classification performance of the S-AAF model.

Results: Experimental evaluations demonstrate that the Swarm Optimized S-AAF model achieves superior predictive performance. It effectively identifies hidden patterns in student data and significantly outperforms existing state-of-the-art methods and standalone optimization algorithms in terms of accuracy and computational efficiency.

Conclusions: The integration of PSO and GA with the S-AAF classifier results in a powerful predictive model that addresses the complexity of educational data. The Swarm Optimized S-AAF model offers a promising approach for improving academic performance prediction and supports more informed decision-making in educational settings.