A model for early prediction of chronic kidney disease using machine learning techniques

Abstract

Chronic Kidney Disease (CKD) is a progressive worldwide health crisis, characterized by silent loss of kidney function that usually goes unnoticed until the disease reaches an advanced stage. Early diagnosis leads to fast treatment, which could help stop the progress of disease and reduce associated mortality rates. Objective of this study was to make an accurate model using machine learning (ML) methods for the early prediction of CKD. The study used a dataset that contains 1,659 records, each of which includes 54 attributes. ML algorithms like K-Nearest Neighbors (KNN), Naive Bayes, Random Forest, support vector machine and Decision Tree classifiers were acted on to predict the likelihood of getting CKD in early stages. Then, these algorithms are integrated as a new Hybrid Ensemble Learning Model (H-EM) to perform better by exploiting their own strengths. The H-EM was found to predict early-stage CKD with excellent performance, % Accuracy, 97% Precision, 96% Recall 97% and an F1 score 96% of this model has a perfect score in all evaluation metrics, making it reliable and robust as is now the state of art for early-stage CKD prediction. In this study, H-EM has illustrated a potential to perform early CKD prediction as it achieves perfect accuracy and performance metrics in early-stage. Given these superior results, we suggest integrating the H-EM in to clinical decision support systems for improve early detection of CKD. Future research should test this model on bigger and more diverse datasets to evaluate its generalizability.