Abstract:
Chronic kidney disease (CKD) is a serious and common health issue affecting millions worldwide. Early detection and treatment of CKD can prevent or delay the need for dialysis or transplantation, which can significantly improve patient outcomes. In this study, we used the UCI data repository to classify the risk of CKD using machine learning techniques. We balanced the dataset using the ADASYN technique to ensure that it was representative of the population. We then evaluated the performance of three algorithms: Random Forest, Naive Bayes, and CatBoost. Our results showed that all three algorithms had high accuracy, with Random Forest and Naive Bayes achieving 99.60% and CatBoost achieving 99.21%.
Additionally, all three algorithms had a precision of 1, and the highest recall value was 99.23% for Random Forest. The F-1 score was also highest for Random Forest at 99.60%. Finally, the ROC_AUC score was 1 for all three algorithms, indicating that they could effectively distinguish between high and low-risk individuals. These results suggest that machine learning can be a powerful tool for classifying the risk of CKD. Further research is needed to validate these findings and to develop more advanced machine-learning techniques for the early detection and treatment of CKD.
