Deep Learning Models for Air Pollution Forecasting in Seoul South Korea

Air pollution has been a major cause of health problems in several countries such as South Korea which is a country with rapid industrial and population growth, it urges the government to pay more attention to this issue. Due to the harmful effects of air pollution, many researchers conduct studies to predict the air quality index as an effort to prevent more severe health issues. In this paper, we propose three deep learning models, namely: Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and combined CNN- LSTM to do air pollution forecasting. We mainly focus on the performance of the models applied in the time-series forecasting task as a supervised learning problem. We use the data from Seoul Metropolitan Government collected hourly from 2017 to 2019 at some stations. The experiment was carried out on sulfur dioxide (SO2), carbon monoxide (CO), nitrogen oxide (NO2), ozone (O3), and two particulate matter (PM) concentrations. We evaluate our model using root mean squared error (RMSE) to compare the models’ performance. The result shows that with normalization CNN model gives the lowest RMSE value, however without normalization the combined CNN-LSTM gives the lowest RMSE value. It proves that the model can predict the air quality index in Seoul South Korea.