Diabetes Mellitus Prediction using Classification Techniques

Abstract

Diabetes is a metabolic disease affecting people in almost every country and it may lead to severe problems like stroke, kidney failure or premature death if it is not predicted at the early stage. To mitigate this many researchers are working to predict the diabetes at early stage using several methods. Different accessible conventional techniques are carried out to diagnose diabetes depend on physical and substance tests. Several d ata mining methods were designed to overcome these uncertainties. Classification techniques like Decis ion Tree, K Nearest Nei ghbors, and Support Vector Machines are used to classify the patients with diabetes mellitus. The performance of the se applied techniques are determined using the factors precision, a ccuracy, Sensitivity, and Specificity. The results obtained proved that SVM outperforms decision tree and KNN with highest accuracy of 90.23%. Performance analysis of these classification methods helps us to decide which appropriate technique to choose in future for analysing the given dataset