User-cloud-based ensemble framework for type-2 diabetes prediction with diet plan suggestion

Abstract

Currently, many individuals are experiencing diabetes, which is attributed to work-related stress and unhealthy lifestyles. Often, people are only aware of their health status once symptoms manifest or they undergo a medical examination to obtain a diagnosis. However, the conventional machine learning methods failed to classify diabetes in the initial stage. Therefore, this work implements the user-cloud-based ensemble framework for type-2 diabetes prediction (DP-UCE). Initially, the Pima Indian Diabetes (PID) dataset is considered. Here, the cloud application trains the PID dataset using an ensemble model. Here, the dataset preprocessing is performed to remove missing values for unknown symbols. Then, three individual models, such as a decision tree classifier (DTC), support vector machine (SVM), and artificial neural network (ANN), were trained with the preprocessed dataset. Then, the bagging ensemble classifier is trained with the DTC, ANN, and SVM features. In addition, the user application is initialized, and type-2 diabetes status is predicted from the uploaded test data of the user using a bagging ensemble classifier. Moreover, based on prediction results, the user is also given a diet plan. The simulations conducted on the PID dataset show that the ensemble classifier resulted in 87.41 % accuracy, which is superior to state-of-the-art methods.