Cloud of Things (CoT) based Diabetes Risk Prediction System using BiRNN

Abstract

The introduction of Internet of Things (IoT) technology witnesses the continuous and distributed connectivity between different objects and people. Currently, with the emerging technological advances, IoT integrates with the cloud and evolves into a new term called “Cloud of Things” to further enhance human lives. Using predictive analytics and Artificial Intelligence (AI) approaches in the healthcare area allows for the development of more reactive and smart healthcare solutions. As a subfield of AI, the Deep Learning (DL) technique has the potential to analyse the given data accurately, provide valuable insights, and solve complex challenges with its ability to train the model continuously. This study intends to implement a deep learning model – Bidirectional Recurrent Neural Networks (Bi-RNN) to obtain a timely and accurate prediction of diabetes risk without requiring any clinical diagnosis. This method of processing the time series data will highly assist in ensuring preventive care and early disease intervention. The proposed model collects real-time data from IoT devices along with the medical data stored in Electronic Health Records (EHR) to perform predictive analytics. The proposed Bi-RNN based diabetes prediction model results in an accuracy of 97.75%, which is comparatively higher than other existing diabetes risk prediction models.