A Study on Different Hybrid Deep Learning Approaches to Forecast air Pollution Concentration of Particulate Matter

Implementation of sophisticated and hybrid technology for air pollution concentration forecasting of particulate matter is considered to be a critical issue nowadays due to rapid urbanisation and industrialisation leading to higher air pollution levels. The PM impacts human health critically, increasing morbidity rates globally. The paper aims at studying different hybrid, deep-learning approaches to forecasting the air pollution concentration of PM. adopting a secondary, qualitative and deductive approach, the paper conducts comprehensive data collection and analysis. The results indicate that using multiple network layers within a deep learning hybrid model holds the most accuracy in practical usages such as CNN-LSTM model and GRU/Bi-GRU modelling. It is concluded that the hybrid deep learning approach enables the identification of hidden patterns in the PM concentration that can be used for adopting mitigation approaches. However, issues related to financial, logistical and infrastructure related issues are identified in the implementation process of such hybrid models widely, especially in low- and medium-income countries.