Air quality forecasting using deep learning and transfer learning: A survey

• Main Authors: Yang, Junzi, Ismail, Ajune Wanis
• 格式: Conference or Workshop Item
• 出版: 2022
• 主題: QA75 Electronic computers. Computer science
• 在線閱讀: http://eprints.utm.my/id/eprint/98852/; http://dx.doi.org/10.1109/GlobConPT57482.2022.9938230

Because of the air pollution problem, if we can dig out the change rule of air quality from historical data or related regional data, we can predict the future development trend of air quality in advance and do an excellent job of preventing air pollution problems. It not only provides reliable help for environmental protection departments to control air pollution, but also provides a reference for public travel. This paper reviews some of the latest research methods in air quality prediction, including hybrid deep learning model and transfer learning model. The hybrid deep learning algorithms are mainly studied from the aspects of traditional statistical methods, machine learning, deep learning, ensemble learning, attention mechanism and optimization algorithm. This paper reviews the application of transfer learning method in air quality prediction from three aspects of pre-training and fine-tuning, multi-source transfer learning and other methods, which is also novel in this paper. Finally, the advantages and disadvantages of hybrid deep learning and transfer learning algorithms are analyzed, which provides a direction for air quality prediction research.