Implementation of an Air Quality Monitoring System and Prediction Using LSTM Ensemble Modeling for the Availability of Multi-use Facilities of Health-vulnerable Groups: a Preliminary Study

• Published in: Quantitative bio-science Vol. 43; no. 1; pp. 11 - 22
• Main Authors: Kwang Hyeon Kim, Hae-Won Koo, Jae-Keun Cho, Je-Hyeon Song, Se-Hyeong Jeon
• Format: Journal Article
• Language: English
• Published: 계명대학교 자연과학연구소 01.05.2024; 자연과학연구소
• Subjects: 통계학; Air quality prediction; LSTM; Health-vulnerable group; Air quality monitoring; Multi-use facility
• ISSN: 2288-1344; 2508-7185
• DOI: 10.22283/qbs.2024.43.1.11

This study aims to ascertain the implementation of an air quality monitoring system and artificial intelligence modeling for predicting air quality in a multi-use facility setting. We designed and deployed a system with sensors for temperature, humidity, TVOC, CO, CO2, PM10, and PM2.5 in a wireless real-time environment within a multi-use facility, collecting 4373 sensor measurements during a month-long trial. We conducted LSTM ensemble modeling and predictions for each of the seven sensor measurements and calculated the RMSE, MAE, MAPE, accuracy, and loss metrics to measure the model performance. The monthly average values of TVOC (1853.54±2859.28, Q1: 168, Q3: 2114), PM10 (9.42±6.89), and PM2.5 (8.93±6.01) were analyzed. Monitoring the air quality of a multi-use facility for a month revealed daily and weekly periodicity in TVOC, CO2, PM10, and PM2.5 through seasonal decomposition analysis. Predictions were performed using the LSTM ensemble network, and the MAPE of the temperature and humidity were analyzed as 0.04 and 0.06, respectively. This preliminary study confirmed that the developed system and AI model, which measure and predict air quality, can be deployed in hospitals and other multi-use facilities that cater to vulnerable groups. KCI Citation Count: 0