Diagnosis of Chronic Kidney Disease using effective classification and feature selection technique

The massive amount of data collected by healthcare sector can be effective for analysis, diagnosis and decision making if it is mined properly. Hidden information extracted from the voluminous data can provide help and remedy to handle critical healthcare situations. Chronic kidney disease is a fatal illness of kidney which can be prevented with early correct predictions and proper precautions. Data mining of the information collected from previously diagnosed patients opened up a new phase of medical advancement. However, specific techniques must be executed to accomplish better consequence. In this manuscript the capability of the classification of Support Vector Machine, Decision tree, Naïve Bayes and K-Nearest Neighbor algorithm, in analyzing the Chronic Kidney Disease dataset collected from UCI repository, was investigated to predict the presence of kidney disease. Data set has been analyzed in terms of accuracy, Root Mean Squared Error, Mean Absolute Error and Receiver Operating Characteristic curve. In the present study, Decision tree shows promising results when implemented through WEKA data mining tool. Ranking algorithm provides vital improvements in classifications with proper number attributes. 15 proves to be the magic number for selecting attributes for the given dataset resulting highest percent of improvement in accuracy.