Data Mining Classifier for Predicting Diabetics

Abstract

Diabetes mellitus is very serious health problem today. If the disease is not identified and treated it can cause very dangerous health hazards. It is very important to predict the diabetic condition to control it through proper medical treatment. There are various hardware and software methods used to predict and classify the data. Various classification algorithms are used to classify the status of a person that whether a person is diabetic or not using various physical and biochemical characteristics. Different methods have different techniques and varying accuracy. In the present work a method for classifying the diabetic data is developed. It is the KNN classifier with six neighbor instance to classify any data record. KNN is lazy learning algorithm which classifies the data based on the majority of neighbors. For the better and accurate classification purpose normalization technique is used to normalize the data. Normalization is a pre processing technique that normalizes the domain value of any variable between 0 and 1. PIMA Indian dataset is used in this study, which include 768 records and 9 parameters.  After classifying the data set using KNN algorithm, Cross validation method is used to compute the accuracy of the algorithm.  The present work is compared with various existing method and it shows a significant enhancement in result by providing 100% accuracy which calculated through cross validation method.