Improved Prediction of Diabetes Mellitus using Machine Learning Based Approach

The diabetes is one of the most commonly occurring chronic diseases in human being. Statistical models are availabel for prediction of diabetes but these provide poor performance. This article proposed machine learning based model for prediction of diabetes disease. Three supervised machine learning algorithms namely K-NN, Linear SVM and Random Forest have been chosen for diabetes prediction for early diagnosis. The area under the curve and accuracy of each of these models have been obtained using PIMA Indian Diabetes dataset from UCI repository. The comparative results demonstrate that among these three algorithms random forest is the best model in terms of accuracy of 78.57 and AUC of 95.08 for diabetes risk prediction. The contribution of this article will help the healthcare professionals for the early prediction of the disease and taking appropriate treatment. The proposed approach can be applied for detection of other diseases.