ABSTRACT
This paper investigated the influence and interactions of air pollution concentrations by using the stochastic boosted regression trees between variables for each station and the impact of the COVID-19 Movement Control Order at Ipoh City air quality station. The one-hour data were gathered from the Department of Environment from January until June 2019 and 2020. Two thousand two hundred thirty-one data of particles, gases (Nitrogen oxides, Sulphur Dioxide, Ozone, Carbon Monoxide) concentrations and meteorological data (wind speed, wind directions, temperature, and relative humidity) were captured. The BRT model development process with an algorithm using a comprehensive package, R Software and its packages to understand the variability and trends. It was found that the relationship between the number of samples and number of trees (nt) of 4372 for oob were found the best iterations obtained. The performance of the boosting model was assessed and found that the FAC2 was 0.91, the R2 values were above 0.56 (R = 0.74), and the Index of Agreements (IOA) was 0.67, which fall ranges are within an acceptable for model performance. The Relative Variable Importance (RVI) that influenced PM2.5 for non-MCO data was CO (18.9% ), SO2 (14.6 %), O3 (12.9 %), and wd (10.66 %) while CO (22.6%), RH (13.4%), 14.7% and O3 (12.1%) were RVI factors influenced to PM2.5 concentrations during MCO periods. Estimating the strength of interaction effects (SIE) between variables was 0.24 for CO-wind directions, followed by 0.19 for ozone-wind speeds and 0.15 for NO2-CO. Results showed that the model developed was within the acceptable range and could be used to understand particles and identify important parameters that influence particle concentrations.