Performance Analysis of Classifier for Chronic Kidney Disease Prediction Using SVM, DNN and KNN

Abstract

AbstractDisruption of the regular operation of the kidney is named chronic kidney disease (CKD). CKD is widespread, and the death rate due to this increases rapidly. To reduce the amount of death, early detection of CKD is necessary. This paper aims to help medical practitioners to diagnose CKD patients by applying machine learning (ML) techniques. We have applied several ML algorithms to the chronic kidney disease dataset which is archived at the machine learning repository of the University of California Irvine (UCI). The classification approaches have been analyzed in this study including deep neural network(DNN), support vector machine (SVM) and K-nearest Neighbor (KNN). To fulfill this study, the missing values have been imputed with different techniques according to the characteristics of the features and relations among them. Hyperparameters of each algorithm have been tuned through experiments. The proposed approach has been evaluated with the best-tuned parameter. The assessment has done based on different performance metrics such as train–test sensitivity, accuracy, f-measure, specificity and Matthews correlation coefficient (MCC). The empirical result shows that SVM and KNN have enhanced accuracy, and DNN shows the most optimistic result with 100% accuracy compared to the existing.KeywordsChronic kidney diseaseSupport vector machineDeep neural networkK-nearest neighbors