Missing Data Classification of Chronic Kidney Disease

Abstract

In this paper we propose an approach on chronic kidney disease classification with the presence of missing data. We implemented a classification system to solve the challenge of detecting chronic kidney diseases based on medical test data. The approach is comparing three different techniques that deals with missing data including deletion, mean imputation, and selection of best features. Each techniques is tested using the K-NN classifier, Naïve Bayes classifier, decision tree, and support vector machines (SVM). The final accuracy of each system is determined using 10-fold cross validation.