Development of Deep Learning-Based Mobile Application for Predicting Diabetes Mellitus

Abstract

With the growing demand for intelligent services on mobile devices, deep learning-based mobile applications are expected to progress even further in the coming year. In the advent of this technology, a deep learning model embedded in a mobile application can play a vital role in predicting a certain kind of disease like diabetes mellitus. Many studies have been performed in the past few years to predict diabetes mellitus using various algorithms of machine learning and deep learning. However, these researches are mostly focused on the development of the predicting model. This study aimed for developing a mobile application that is deep learning-based for predicting diabetes mellitus. Using the TensorFlow platform, the Sequential function was used in building the diabetes prediction model. The model was then transformed into a ‘tflite’ format which was deployed in the development of mobile application using the Android Studio integrated development environment (IDE) to predict if a person has diabetes mellitus. The deep learning model demonstrated considerable accuracy of 93%. Additionally, the application also provides some important instructions for the end-users and facts about diabetes mellitus. The developed deep learning-based mobile application is an important new technology for diabetes mellitus early detection. If the prediction is positive, the lifestyle could change, and a serious complication will be avoided.