Modeling Fluctuation of PM10 Data with Existence of Volatility Effect

摘要

Modeling time series data of particulate matter (PM) will provide a good understanding about the dynamic behavior of this pollution variable. In fact, a suitable model can be used as a practical tool for planning purposes and controlling adverse effects of air pollution. This article utilized an autoregressive integrated moving average (ARMA) with the combination of generalized autoregressive conditional heteroscedastic (ARCH/GARCH) to provide a suitable model that can overcome the problematic volatility effect that exists in the PM10 data. Hourly PM10 data for the city of Kuala Lumpur have been analyzed. Based on several statistical approaches, such as the autocorrelation function, R-2 coefficient, and Akaike's Information Criterion, an ARMA(1,0)-GARCH(1,1) has been determined to be the best model to describe the data. In fact, incorporation of GARCH(1,1) is able to improve forecasting performance of PM10 data, instead of relying on only a single ARMA(1,0) model.