Prediction of Ground Level Ozone by applying Artificial Neural Network nearby School Area in South Seberang Perai, Penang and Parit Buntar, Perak, Malaysia
Ground level ozone gives intensively concern of its adverse effects towards human health and environmental. Increasing numbers of vehicles which also known to be one of the sources of its precursors that emitted from vehicle exhausts leads to the production of ground level ozone. Meteorological conditions are also affecting the production of this pollutant. Feedforward Backpropagation models were developed by using artificial neural network to predict ozone concentrations nearby selected schools area by considering the relationship with its precursors and meteorological parameters. From the results, the best hidden nodes found for all FFBP models for SSH, SSN, SST and STR were 17, 12, 15 and 14, respectively. SSN found to has the best FFBP models with the highest accuracy measures of IA, PA and R2 which are 0.9832, 0.9686 and 0.8674, respectively if compared to other schools. SSN also recorded the lowest error for RMSE with the value of 3.3327. However, STR found to has the lowest error of NAE with the value of 0.1004. Nevertheless, the value of NAE for SSN can still be considered as lower amongst all schools.
Keywords- ozone precursors; hidden node, transfer functions; training algorithm; feedforward backpropagation