An Experimental Evaluation of Hybrid Learning Methodology based Internet of Things Assisted Health Care Monitoring System
Abstract In this modern world, every individual uses intelligent devices to lead a day-to-day activity intelligently. Using the latest technologies such as deep learning, the Internet of Things (IoT) forth provides standard prediction and communication abilities to the existing applications to properly provide rich support to the clients. Many commercial and non-commercial organizations almost adapt these technologies to modify their physical records digitally. This paper designed a novel health care monitoring scheme by adapting these technologies to provide an intelligent monitoring system to analyze patients over random instances with periodic intervals. This paper introduced a new learning-based scheme called Deviated Learning-based Health Analysis (DLHA), in which it combines the conventional algorithms such as Convolutional Neural Network (CNN) and the Support Vector Classification (SVM) logic in a transparent manner. The logical evaluations of the proposed approach called DLHA assessed by extracting the layers from the CNN, appending the classification logic of SVM into the CNN layers, and defining a new algorithm to predict patient health intelligently. The association of sensor-based smart device called Smart Health Indicator (SHI) provides significant support to the proposed approach with the association of intelligent sensors such as Heartbeat Analyzer, Body Temperature Estimation Sensor, Breath Sensor, Global Positioning System (GPS), and the useful Internet of Things enabled controller called ESP8266. Using this SHI kit, the patient details are monitoring instantly and reporting it to the remote server periodically to analyze the health summary without any interventions. The proposed deep learning strategy called DLHA acquires the data from the intelligent health care kit SHI and processes it using classification principles. The records collected from the kit were manipulated according to the process of the trained model generated from the previous testing samples of the patients. The dataset used in this system is generated dynamically from the real-time patient health record and processes the testing report of the patient accordingly. The processed record is appended into the dataset for further reference. The resulting section provides proper proof of the efficiency of the proposed approach in a transparent manner with graphical representations. For all this system is more significant to identify and monitor the health details of the patient in clear manner with proper specifications.