Modelling of Surface Ozone via Neural Network and Statistical Approaches

Abstract

Surface ozone is one of the air pollutants that contribute to the air pollution. Its sources could be from anthropogenic activities and natural disasters during the past few decades. The purpose of this study is to determine the most appropriate model for forecasting the surface ozone at Shah Alam, Selangor and Larkin, Johor. Several analytical model such as Time Series Regression (TSR), Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) are fitted to the ozone concentration. Model comparison is carried out via performance indicators (RMSE, MSE, R square). The results show that ANN provides better performance compared to TSR and MLR. Between the two stations, Larkin, Johor provides high accuracy in forecasting surface ozone concentrations for each model with minimum MSE (0.000009), RMSE (0.0042) and high value of R2 (0.33) compared to the station in Shah Alam, Selangor.