Statistical Analysis and Predictive Modelling of Air Pollutants Using Advanced Machine Learning Approaches

Abstract

Air quality forecasting is important as it provides early warning of air pollutants, which could be harmful to human health and the environment. This study presents a statistical analysis and prediction of various air pollutants (such as PM 10 , CO, SO 2 , NO 2 and O 3 ) and the air pollution index (API) in Labuan, Malaysia using advanced machine learning approaches. The exponential triple smoothing (ETS) and seasonal autoregressive integrated moving average (SARIMA) forecasting methods were used in the study. Air pollutants and API data from 2000 to 2018 were analyzed and tested with these forecasting models. The ETS model with 68% confidence interval (CI) fitted well for all air pollutants except SO 2 , which was best fitted by the 95% CI model. The SARIMA (1, 1, 1)(0, 1, 1)12 model was found to be the most appropriate for forecasting different air pollutants and API in Labuan. The ETS model was more suitable for forecasting CO and SO 2 and the SARIMA (1, 1, 1)(0, 1, 1)12 model was more suitable for forecasting PMio, NO 2 , O 3 and API. The ETS model predicts that the annual concentrations of CO and SO 2 in 2030 would be about 0 ± 0.81 ppm and 0 ± 0.002 ppm, respectively, at 68% CI. The SARIMA (1, 1, 1)(0, 1, 1)i 2 model predicts that the annual concentrations of PMio, NO 2 and O 3 in 2030 would be about 38.7 ± 14.8 pg/m3, 0 ± 0.006 ppm and 0.01 ± 0.008 ppm, respectively, and API value of about 44 ± 15 in Labuan.