Prediction performance of individual and ensemble learners for chronic kidney disease

Automating the process of predicting diseases prove assistive and time-saving for a practitioner in the field of medical diagnosis. The accurate prediction of any disease not only helps the patients know about their health but also helps the doctors in medication suggestion well in advance. In today's lifestyle, advance knowledge about health and proper care can add a number of living days to a patient's life. In this paper, the prediction of chronic kidney disease (CKD) is performed using individual and ensemble learners. The experiments are performed on CKD dataset was taken from UCI repository. The three different classifiers from individual classifiers, namely, Naive Bayes(NB), minimal sequential optimization (SMO), J48, and three ensemble classifiers, namely, Random Forest (RF), bagging, AdaBoost respectively are used for prediction. We have used the open source, weka tool, for all the experiments. The results are evaluated using accuracy, precision, recall, F-measure and ROC performance measures. The results suggested that the decision tree based individual learner (J48) and random forest from ensemble classifier respectively perform better than the other classifiers.