The nonlinear autoregressive exogenous neural network performance in predicting Malaysia air pollutant index

• Main Authors: Rosminah Mustakim, Mazlina Mamat
• Format: Article
• Language: English; English
• Published: UniSE Press 2021
• Subjects: TD878-894 Special types of environment Including soil pollution, air pollution, noise pollution
• Online Access: https://eprints.ums.edu.my/id/eprint/32936/1/The%20nonlinear%20autoregressive%20exogenous%20neural%20network%20performance%20in%20predicting%20Malaysia%20air%20pollutant%20index.pdf; https://eprints.ums.edu.my/id/eprint/32936/4/The%20nonlinear%20autoregressive%20exogenous%20neural%20network%20performance%20in%20predicting%20Malaysia%20air%20pollutant%20index%20_ABSTRACT.pdf; https://eprints.ums.edu.my/id/eprint/32936/; http://tost.unise.org/pdfs/vol8/no3-2/ToST-CoFA2020-305-310-OA.pdf

Predicting the air quality is important particularly in the areas where air pollution is becoming a major health problem. This paper presents and evaluates the Nonlinear Autoregressive Exogenous (NARX) Neural Network performance in predicting the Air Pollutant Index (API) at three industrial areas in Malaysia: Pasir Gudang, Larkin and TTDI Jaya. The NARX was implemented in an open loop feed-forward architecture and was trained to produce an hour ahead API prediction based on the past values of air quality and meteorological parameters. Six air quality parameters: CO, NO2, O3, PM2.5, PM10, SO2, and three meteorological parameters: wind direction, wind speed and ambient temperature were used as input while the API was set as the output. The prediction performance was measured by using the Coefficient of Determination (R2) and Root Mean Square Error (RMSE) tests. Results show that the performance of NARX model was encouraging with R2 value above 0.97 and RMSE value around 1.21 based on the data collected in 2018 at the three monitoring stations.