Prediction of Air Quality Using LSTM Recurrent Neural Network

The rapid increase of industrialization and urbanization significantly draws the interest of researchers towards the prediction of air quality. Efficient modelling of air quality parameters using deep learning methods can facilitate the imminent implications of air pollution. However, existing methods weaken at consideration of long-term dependencies for multiple parameters. The present study aims prediction of air quality of New Delhi based on concentration of multiple parameters, namely PM2.5, PM10, CO, O3, NO2, and SO2. The study uses long short-term memory (LSTM) approach due to its efficiency over other deep learning methods and referred it as A-LSTM prediction model. It supports multiple layers to add more linearity to the desired output. Performance of A-LSTM is evaluated for prediction of year 2019 data. Mean absolute error, root mean squared error, precision, recall, and F1-score metrics are considered for comparison with other three prediction models, namely support vector regressor (SVR), SVR with LSTM, and I-LSTM.