KIDNEY DISEASE DETECTION IN INDIAN PATIENTS IN AN EARLY STAGE USING WEKA TOOL

Abstract

Systems like health care, educational organization, financial institution etc., and produce huge amount of data. For decision making, there is a need of processing a raw data generated by these systems automatically. For example there is need of decision support system in health care to decrease the hospitalization rate. The quality of service provided will be low if the rate of hospitalization is more [1]. Machine learning (ML) is involved in the automatic identification of hidden patterns if there exists any in a given raw data. ML helps discovering knowledge from within a raw data that data mining aims at. Such knowledge plays a vital role in decision making. Mathematical and statistical techniques, namely, regression, classification, clustering, and association rule mining are generally employed [2]. Several software are available. We here discuss features of one such, WEKA, a collection of ML tools for data pre-processing, classification, regression, clustering, association rules, and visualization. The WEKA library can be used directly or one can embed the functions in Java code. The case presented here is from the health care sector: Diagnosis of chronic kidney disease based upon the patient’s history. Record containing blood pressure, sugar, red blood cells count etc., of 400 patients has been used to build a classification model [3]. Results are compared with LMT(Logistic model tree), Random Tree, REP Tree (Reduced Error Pruning) and the accuracy is more in J48. In LMT method the results are 98%, in Random tree it is 96.5% , REP Tree method results are 96.75%whereas with J48 algorithm the result is 99%. Classification result of Support Vector Machine (SVM) and Artificial Neural Network (ANN) algorithm shows the accuracy of 76.32% and 87.70% respectively [4]. The paper concludes that J48 classification algorithm comparatively generates the best result.