An innovative framework for student adaptability prediction using discriminant-based classification, feature selection, and sensitivity analysis

Abstract Accurately predicting student adaptability is essential for enhancing academic performance and personal development. The present study presents a classification-based approach utilizing Quadratic Discriminant Analysis (QDA), Extreme Gradient Boosting (XGBoost), and Linear Discriminant Analysis (LDA) to predict student adaptability. To enhance model performance, the Turbulent Flow of Water-based Optimization (TFWO) and Giant Armadillo Optimization (GAO) algorithms are employed. The dataset, consisting of 1,205 student records with demographic, educational, and socio-economic attributes, is split into training and test sets. The two best-performing hybrid models undergo a sensitivity analysis using the FAST approach. Experimental results demonstrate that the QDA model optimized with GAO (QDGO) achieves the highest classification accuracy of 0.975, outperforming other model configurations. These findings underscore the potential of optimized classification models in student adaptability prediction, providing valuable insights for educational institutions. Future studies can explore additional hybrid models and larger datasets to further refine predictive accuracy.