A Framework for the prediction of Diabtetes Mellitus using Hyper-Parameter tuned XGBoost Classifier
R Gayathri, Peeta Basa Pati, Tripty Singh, Rekha R Nair
Abstract
Diabetes Mellitus is caused by the increase of blood glucose level in the body. It is likely to be the most common disease for the human community which is about 8.8% of the entire population. It is important to identify this disease in the initial stages of the development to reduce the risk factor of further heart diseases and other vital problems. Machine Learning is the one of the recent developments which is very helpful in clinical care guidelines. The PIMA Indian Diabetes dataset is used to predict Type-2 Diabetes Mellitus based on certain clinical diagnostic measurements for females. In this paper, we proposed a framework for prediction of Diabetes Mellitus using Optimised Gradient Descent Boosting Classifier. The performance metrices such as Accuracy, Sensitivity, Specificity and F1 scores are chosen. These experiments are conducted for PIMA Indian Diabetes Dataset and the proposed classifier yields 94.5 %, 92.4 %, 96 % and 92 % for the values in Accuracy, Sensitivity, specificity and F1 score and also compared with other classifier’s like K-NN, QDA, SVM and etc.