Identification of Nominal Attributes for Intelligent Classification of Chronic Kidney Disease using Optimization Algorithm

Abstract

Diagnosis of chronic kidney disease (CKD) has mainly significant in the medical field of data mining. The objective of this paper is to predict chronic kidney disease using only nominal attributes which results compared to numerical and both nominal and numerical attributes. The classification and prediction of this paper presents correlation based feature selection (CFS) technique applied to extract important attributes and classifying them into CKD and not CKD. The CFS approach is applied on the data types of nominal, numerical and both nominal and numerical attributes in feature selection. The result of the CFS approach compared to three ranker approaches such as Information Gain, Gain Ratio, and ReliefF approach for feature selection. The accuracy of correlation based feature selection and sequential minimum optimization (CFS-SMO) approach achieved 98.5% for nominal, 95.25% for numerical and 98.5% for nominal and numerical. These experimental results declare that the correlation based feature selection (CFS) successfully extracted features from the benchmark approval of own and original chronic kidney disease (CKD) dataset and SMO classified the status of kidney disease. Therefore CFS-SMO considered as an optimistic tool to diagnosis kidney disease accurately which supports medical experts to make decision correctly.