Enhanced Feature Selection for Chronic Kidney Disease Detection: A Hybrid Integration of Simulated Annealing and Recursive Feature Elimination

Chronic kidney disease (CKD) poses a significant global health challenge due to its often asymptomatic early stages, making early detection crucial for effective treatment. This chapter delves into the complexities of feature selection in medical data, highlighting the importance of reducing data dimensionality to improve model performance and avoid overfitting. The proposed hybrid method combines simulated annealing (SA) and recursive feature elimination (RFE) to achieve a more accurate and efficient feature selection process. SA, inspired by metallurgical annealing, explores a broad feature space to identify potential candidates, while RFE refines these subsets by iteratively eliminating the least important features. This dual approach ensures a comprehensive and precise selection of features, enhancing the predictive accuracy of feedforward neural networks (FNNs) used for CKD detection. The chapter also reviews existing literature on CKD detection methods, comparing various feature selection techniques and their impacts on model performance. Through rigorous evaluation metrics, the proposed SA-RFE hybrid method demonstrates superior accuracy, precision, recall, and F1-score, outperforming baseline models and other hybrid approaches. The results underscore the potential of this innovative method to revolutionize CKD detection, offering a more reliable and efficient diagnostic tool.