Proposal of a Methodology for Prediction of Indoor PM2.5 Concentration Using Sensor-Based Residential Environments Monitoring Data and Time-Divided Multiple Linear Regression Model

Abstract

This study aims to propose an indoor air quality prediction method that can be easily utilized and reflects temporal characteristics using indoor and outdoor input data measured near the indoor target point as input to calculate indoor PM2.5 concentration through a multiple linear regression model. The atmospheric conditions and air pollution detected in one-minute intervals using sensor-based monitoring equipment (Dust Mon, Sentry Co Ltd., Seoul, Korea) inside and outside houses from May 2019 to April 2021 were used to develop the prediction model. By dividing the multiple linear regression model into one-hour increments, we attempted to overcome the limitation of not representing the multiple linear regression model’s characteristics over time and limited input variables. The multiple linear regression (MLR) model classified by time unit showed an improvement in explanatory power by up to 9% compared to the existing model, and some hourly models had an explanatory power of 0.30. These results indicated that the model needs to be subdivided by time period to more accurately predict indoor PM2.5 concentrations.