An Optimized Deep Belief Neural Network Approach for Chronic Kidney Disease Prediction Using GWO

Abstract

In today’s healthcare system, accurate classification of medical data is crucial for effective disease diagnosis and prediction. Machine learning (ML) and deep learning (DL) models are extensively utilized for these tasks, yet they often face challenges related to computational complexity and processing efficiency. This research introduces a novel model for diagnosing chronic kidney disease (CKD) that integrates advanced data preprocessing, feature selection, and classification techniques. Missing data is addressed using Multiple Imputation by Chained Equations (MICE) to maintain dataset integrity, followed by feature selection using Grey Wolf Optimization (GWO) to enhance prediction accuracy. Finally, a Deep Belief Neural Network (DBNN) performs the classification. Model effectiveness is evaluated through sensitivity, accuracy, and specificity, demonstrating superior performance compared to existing methods.