Type 2: Diabetes mellitus prediction using Deep Neural Networks classifier

Abstract

Diabetes is a severe disease, most of the people are not aware of the risk associated with the disease because of that people die due to diabetic nephropathy, cardiac stroke and some other disorders. Therefore, early identification of diabetes helps to maintain sound health and life. Deep Learning approaches are used to predict diabetes accurately as humans do. In this paper, Deep Neural Network (DNN) classifier, an unsupervised learning approach is used for accurate prediction on Pima Indian Diabetes dataset and Feature Importance model that is bagged with Extra Trees and Random Forest is used for feature selection. The Pima Indian Diabetes dataset (PID) was acquired from the repository of UCI. The existing dataset has experimented with different formats of train test splits. The performance of the model was evaluated through accuracy, specificity, sensitivity, recall and precision. The model acheived 98.16% accuracy with random train-test split and it is observed that, the model obtained better performance than other state-of-art methods.