Machine Learning Based Prediction and Modeling in Healthcare Secured Internet of Things

In this paper, we present the concept and the prototype implementation of our novel "Smart Observatory of Involuntary Medical Seizures (SOIMS)". SOIMS merges Wireless Body Area Networks (WBAN), Internet of Things (IoT) and Machine Learning (ML) as an intelligent platform for the prediction and modelling of involuntary seizures. The prediction process is elaborated with our proposed algorithms, namely, Qualifying Linear Regression Algorithm (QuLRA), Selective Clustering Algorithm (SeCA) and Real Time Clusters Correlation Algorithm (RT2CA). The assessment of the proposed system is validated based on the Physionet ECG patients' dataset. The implementation of the prototype involves an IoT/WEB proxy security embedded for translation between nodes CoAP/DTLS protocol and Hospital Information System (HIS) HTTP/TLS protocol. Our proposed solution outperforms existing schemes in the literature at different levels, namely: a) it uses a hierarchical combination of machine learning and prediction algorithms; b) it is open-source, interoperable and user friendly; c) it is a secured prototype implementation; and d) it reaches a higher rate of accuracy according to the correlation criterion.