A Pilot Study using Chaotic Approach to Determine Characteristics and Forecasting of PM10 Concentration Time Series

Forecasting of PM10 concentration is important as breathing air containing PM10 can lead to chronic diseases such as cancer and bronchitis. This study is a pilot study using chaotic approach to forecast PM10 in Malaysia. Studied data is a time series of observed hourly PM10 at benchmark station located in the district of Jerantut in Pahang state. Chaotic approach has two steps, namely the phase space reconstruction and the forecasting process. Through step 1, multi-dimensional phase space is reconstructed using the parameters of the delay time τ = 7 and embedding dimension m = 4, respectively, derived from the average mutual information and Cao method. The results from the phase space diagram and parameter plot of Cao method demonstrates that the data are chaotic. Through step 2, 1 h ahead forecasting for a month PM10 time series is carried out using the local approximation method. Correlation coefficient value between the actual and forecasted data is only 0.5692. However, comparison graphs show that forecasted data are close to the actual data with root mean square error value 7.6814. This demonstrates the suitability of the local approximation method to forecast the time series of PM10 and it’s a positive sign that this chaotic approach is applicable to the time series of pollutants in Malaysia.