Prediction Of Diabetes Mellitus Using Measure Of Insulin Resistance: A Combined Classifier Approach

Abstract

Diabetes is a cluster of diseases that are categorized by hyper glycaemia which is the result of defects in insulin secretion by the pancreas, the action of the insulin over the carbohydrates that we consume, or both of the conditions. Diabetes holding hyper glycaemia is resulting with failure of organs at a longer term rate such as eyes, kidney and heart. Since the disease is creating havoc in the human race it’s important to identify the cause with great accuracy and precision. Using data mining and machine learning algorithms we try to find the accuracy of classifying the same. The diabetes dataset is a binary classification problem and its main objective is to analyze if a patient is affected by the disease or not. We concentrate here on the various classifiers and their accuracy results in identifying the presence and absence of diabetes. The study is conducted on classifiers like Decision Tree, SVM, Logistic Regression, Linear Regression, K- Nearest Neighbor, Random Forest and Naïve Bayes algorithms. An in depth analysis is made on the contribution of the attributes in the classification problem.