Modeling multivariable air pollution data in Malaysia using vector autoregressive model

Abstract

In this study, the vector autoregressive (VAR) model was used to model and forecast the multivariable air pollution data in Klang area. Stationary test, Hannan–Quinn evaluation criteria, Granger causality test, R2 coefficient and Root Square Mean Error (RMSE) measurements have been conducted to get the best model and will be used in forecasting. The VAR (7) model is found to be the best model with the highest R2 and lowest RMSE value recorded for each dependent pollutant variable. Based on the fitted VAR (7) model, the VAR model is able to describe the dynamic behavior of multivariable air pollution data of Klang. Forecasts of up to 12 days ahead were constructed with confidence intervals. The VAR model found to provides good forecast accuracy on the data.