Air Pollution Particulate Matter (PM2.5) Forecasting using Long Short Term Memory Model

Air quality significantly affects human health and climate, good air quality and accurate air quality predictions have become one of the most widespread concerns. In previous studies, the datasets were used only in one of the major cities in the world. The dataset used in this study came from 7 different countries and used LSTM and VAR to predict. From several tests, LSTM can provide the best performance for predicting in the long term. The best RMSE value from LSTM is 22,227 by providing 0.0001 for learning rate, the number of neurons and epochs of 50 each, and tested on testing data within four months. Meanwhile, VAR gave the best results with an RMSE of 16,472 on the first month of testing data by giving a P-value is 24. The results show that the RSME value of the VAR is 36,802, the RSME value of the LSTM is 27,902.