A Novel Hybrid Approach for Diagnosing Diabetes Mellitus using Farthest First and Support Vector Machine Algorithms

ABSTRACT Introduction Diabetes mellitus (DM) is a chronic disease related with abnormally high levels of the sugar or glucose in the blood. Diabetes is due to one of two mechanisms such as inadequate production of insulin and inadequate sensitivity of cells to the action of insulin. The purpose of this study is to diagnose the diabetes mellitus by using data mining algorithms. Methods There are many ways to diagnose the diabetes. One of the methods is data mining algorithms. The use of data mining on medical data has fetched about valuable, important and effective achievements, which can enhance the medical knowledge to make necessary decisions. In this paper, we propose an integrated approach of Farthest First (FF) clustering algorithm and Sequential Minimal Optimization (SMO) classifier algorithm for diagnosing the DM. Farthest first clustering algorithm is used to groups the data in to number of clusters. The computation time was reduced greatly due to shrink the size of dataset. The clustering output is given as input to SVM classifier. It classifies the patients into diabetic and non-diabetic i.e. tested positives and negatives with high accuracy. Result The dataset used for the diagnosis of diabetes includes 768 samples from diabetic patients taken from Pima Indians Dataset. Experimental results show that the proposed integrated approach achieved 99.4% classification accuracy for predicting the diabetes mellitus. Conclusion Experimental results proved that our proposed integrated approach achieved a classification accuracy of 99.4% for diagnosing the patient with diabetic and non-diabetic. The experimental results proved that, hybrid approach of data-mining method could help the doctors to make better clinical decisions for diagnosing diabetic patients.