Prediction Modelling Through Chaotic Approach on Ozone (O3) Pollutant Time Series in Shah Alam City

Abstract

Studies on Ozone (O3) particles and their effect have become a significant concern. Exposure to high concentrations of O3 can cause an adverse reaction in the human respiratory system. Therefore, this study aims to predict the monthly O3 time series in the highly populated area of Shah Alam through a chaotic approach. The phase space plot and Cao methods successfully detected the chaotic behaviour O3 time series data. The three parameters are determined before the prediction process: time delay t, embedded dimension m, and nearest neighbour k. This study used t=1 and the value of m is calculated through the Cao method. The last parameter that needs to be determined is by graph plotting k versus correlation coefficient(cc). The combination of parameters t and m will be used for prediction, and the performance measure will be recorded. The prediction process is done using three non-linear methods, namely LMAM, LLAM and ILLAM. LMAM gives the best prediction value by using the combination of parameter t and the value of the performance measure is 0.8486. Trial-and-error method, m = 6, LMAM gives cc = 0.8977 with k = 17, LLAM gives cc = 0.8033 with k =1, and ILLAM gives cc = 0.8265 with k =24. Mean-top-bottom method, m = 9, LMAM gives cc = 0.8863 with k =10, LLAM gives cc = 0.8841 with k = 6 and ILLAM gives cc = 0.9370 with k = 22. This finding indicates that the O3 time series can be predicted using a chaotic approach and the improved method in determining the value can compute better prediction performance.

Keywords

• ozone
• chaotic approach
• non-linear
• phase space plot