Air Pollution Prediction and Classification with a Hybrid ANN-LSTM Model in Modern Cities: A Comparative Study
Air Pollution Prediction and Classification with a Hybrid ANN-LSTM Model in Modern Cities: A Comparative Study
Authors: M. A. I. Rafi, M. R. Sohan, G. J. Rayhan, M. S. I. Khairul, A. A. Noman, and M. H. Nadid
Abstract— Urban environments and industrial activities contribute to air pollution, a critical issue with significant health impacts. This study focuses on air pollution in urban and industrial environments and proposes a hybrid model to predict and classify air pollution levels. The model combines Artificial Neural Networks (ANN) and Long-Short-Term Memory (LSTM) networks to analyze air quality data’s complex and time-dependent nature effectively. A comprehensive dataset of over 60,000 samples of air pollutants, including particulate matter (PM2.5), was collected to evaluate the hybrid model. These pollutants were carefully selected as key indicators of air quality. The hybrid model’s result was compared to various models such as Linear Regressions, Random Forests, Decision Trees, ANNs, and LSTMs. The results showed that the hybrid ANN-LSTM model demonstrated better performance compared to the other models, in terms of precision, recall, F1-score, and accuracy of 94.87%. It also had the lowest prediction errors across MSE, RMSE, and MAE, as well as the highest R2 value and the lowest MAPE, indicating a good performance in forecasting air quality. This hybrid approach to air quality prediction and classification has the potential to significantly improve the management of air pollution and inform public health decisions.
Keywords— Air Quality Index, Particulate Matter, Artificial Neural Networks, Long Short-Term Memory