Abstract:
This research proposes a feature-weighted ensemble learning model to achieve optimal Chronic Kidney Disease (CKD) classification. The number of people affected with CKD continues to grow rapidly because this condition has become a major health challenge affecting more than 850 million individuals worldwide. The accurate identification of diseases remains fundamental to managing the escalating healthcare system challenges. The proposed study develops an improved high-performance diagnostic model by combining state-of-the-art machine learning algorithms with ensemble methods to enhance diagnostic precision. To detect important clinical markers, the proposed approach includes Recursive Feature Elimination (RFE) and Least Absolute Shrinkage and Selection Operator (LASSO) along with other robust feature selection algorithms. According to the outcome, Hemoglobin and Packed Cell Volume demonstrated higher clinical importance than Serum Creatinine and Specific Gravity. Random Forest, XGBoost, CatBoost, and LightGBM classifiers were used, combined with adaptive weights to enhance the model's performance.The UCI CKD dataset was used for this research, and it originally contained 24 features, but preprocessing together with feature selection reduced it to seven features, which produced an accurate model with 99.75% accuracy and 0.50% standard deviation. The model delivers performance that exceeds standard diagnostic methods to generate a dependable and efficient decision-making tool with explanatory capabilities for clinical use. The model demonstrates wide applicability across healthcare systems, which enables it to revolutionize CKD diagnosis and treatment, particularly in resource-limited healthcare environments.This feature-weighted ensemble learning methodology not only enhances the diagnosis of CKD but also demonstrates that data-driven methods are critical to achieving sustainable development in healthcare systems worldwide.
