Towards Smart Healthcare Teleconsultation

A Secure IoT-Edge-Machine Learning Architecture for Diabetes Data Collection and Prediction

Abstract

The increasing demand for medical consultations in developing countries is one of the largest pressures on healthcare systems. In this study, an IoT-enabled device was proposed for automating the collection of physiological data for diabetes classification and prediction. The collected data was stored on ThingsBoard Cloud for convenience in teleconsultations and reduction of physical visits. The system use edge computing (EC) technology implemented through a Raspberry Pi server and makes use of the ThingsBoard cloud for data monitoring and visualization. However, for data confidentiality, integrity, and secure access, the system includes user authorization, data encryption, and secure transmission mechanisms. Furthermore, machine learning (ML) models based on random forest (RF) and XGBoost were used for predicting diabetes from the collected data. Precision, recall, F1-score, and accuracy analysis revealed that RF performs better than XGBoost, achieving 99% overall accuracy with commendable efficiency on all the above metrics. Therefore, future research work should include increasing system security with better threat detection mechanisms, improving ML models by using hybrid mechanisms, and anonymizing data for compliance with data privacy frameworks for health data to realize smarter, more secure, and more accessible healthcare delivery.