Analysis of Chronic Kidney Disease (CKD) using supervised machine learning classifiers and curve fitting

Abstract

Chronic Kidney Disease (CKD) is a fatal disease and proper diagnosis is desirable. The dataset of CKD has been taken from the UCI repository. Supervised machine learning classifiers have been applied viz Naïve Bayes, Logistic Regression, Decision Table, Random Forest and Random Tree. For each classifier, the result is derived based on (i) without preprocessing (ii) SMOTE with resampling and (iii) Class balancer. From the results, it is evident that Random forest classifier is giving the highest accuracy with 98.93 % in Smote with resampling. The significant parameters for CKD are Sugar Level, Aluminum Level and Percentage of Red blood cells. Optimum curve fitting in case of Sugar level is observed in case of a polynomial of order 3, that in case of aluminium level is found in case of a polynomial of order 2 while that in case of Percentage of red blood cells is noted in a logarithmic curve.