| dc.description.abstract |
Air is an important resource for all living things to live for survival. However, in some area, especially in the capital we have found that the air quality is contaminated with pollution which affects people’s health. Unfortunately, there is still no proper way to deal with the problem of fine dust PM2.5 and this problem becomes a major source of severe environmental air pollution both domestically and internationally. The objective of this research is to propose the fine-tune machine learning models which is able to forecast 7-Day PM2.5 in Bangkok. The model could determine appropriate measures to cope with the haze problem in the future. The Long Short-Term Memory models (LSTM), one of the Deep Learning models, was trained using hourly air pollution data from the Pollution Control Department, Thailand, and The Meteorological Department, Thailand. the experiment results shown that Long Short-Term Memory (LSTM) had the best performance in predictions of PM 2.5 in 7 days. The best results included PM2.5, PM10, Wind Speed, Pressure, Humidity, and Temperature. The model performance values were RMSE 8.47, MAE 6.37 and MAPE 25.19%. This research has improved the efficiency of the model to forecast more accurately by choosing Adam Optimizer. |
|