Web-based Non-contact Edge Computing Solution for Suspected COVID-19 Infection Classification Model
- Authors
- Hwang, Tae-Ho; Lee, Kangyoon
- Issue Date
- Oct-2023
- Publisher
- RIVER PUBLISHERS
- Keywords
- Edge computing; COVID-19 classification; non-contact bio-sensor; artificial intelligence; machine learning
- Citation
- JOURNAL OF WEB ENGINEERING, v.22, no.4, pp 597 - 614
- Pages
- 18
- Journal Title
- JOURNAL OF WEB ENGINEERING
- Volume
- 22
- Number
- 4
- Start Page
- 597
- End Page
- 614
- URI
- https://scholarworks.bwise.kr/gachon/handle/2020.sw.gachon/89624
- DOI
- 10.13052/jwe1540-9589.2242
- ISSN
- 1540-9589
1544-5976
- Abstract
- The recent outbreak of the COVID-19 coronavirus pandemic has necessi-tated the development of web-based, non-contact edge analytics solutions. Non-contact sensors serve as the interface between web servers and edge analytics through web engineering technology. The need for an edge device classification model that can identify COVID-19 patients based on early symptoms has become evident. In particular a non-contact implementation of such a classification model is required to efficiently prevent viral infec-tion and minimize cross-infection. In this work, we investigate the use of diverse non-contact biosensors (e.g., remote photoplethysmography, radar, and infrared sensors) for reducing effective physical contact with patients and for measuring their biometric data and vital signs. We further explain a clas-sification method for suspected COVID-19 infection based on the measured vital signs and symptoms. The results of this study can be applied in patient classification by mobile-based edge computing applications. The correlation between symptoms comprising cough, sore throat, fever, headache, myalgia, and arthralgia are analyzed in the model. We implement a machine learning classification model using vital signs for performance evaluation, and propose an ensemble model realized by fine-tuning the high-performing classification models. The proposed ensemble model successfully distinguishes suspected patients with an accuracy, area under curve, and F1 scores of 94.4%, 98.4%, and 94.4%, respectively.
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - ETC > 1. Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.