Advancing Chronic Kidney Disease Diagnosis and Prediction Using a Stacked Model Approach

Chronic Kidney Disease (CKD) is a significant challenge for global health, leading to approximately 1.7 million deaths every year. Traditional diagnostic methods are increasingly supplemented by advanced machine learning (ML) techniques for their superior accuracy in predicting disease. This study leverages data analysis to identify patterns in CKD data and applies various ML classifiers, focusing on a stacking technique to enhance predictive accuracy. Our main goal was to assess various ML classification methods to identify the best model for CKD prediction. We began with a comprehensive data analysis to uncover patterns and insights within the CKD dataset. This analysis informed the selection and development of our models, including Gradient Boosting (GB), XGBoost, AdaBoost, and a Stacking model. We assessed these models based on training and testing accuracy to identify potential overfitting or underfitting and used multiple evaluation metrics, including precision, recall, F-score, and overall accuracy. Our results show that Gradient Boosting achieved an accuracy of approximately 95.91%, XGBoost reached 97.16%, and AdaBoost attained 95.43%. Notably, our proposed stacking model outperformed the others, delivering an exceptional accuracy of 99.15% on the CKD dataset. This highlights the effectiveness of combining data analysis with advanced ML techniques, demonstrating the stacking model’s potential to improve CKD prediction and diagnostic approaches significantly.