Health and Ecological Risk Assessment-Based Air Quality Prediction Framework Using Ensemble Learning Network with Optimal Weighted Prediction Score

Due to the high intensity of air pollutants in urban areas, people are suffering from more breathing-related health issues. These health effects are prominent in both older and younger people. Several methods and techniques are adopted by the government to tackle the high emission of this air pollutant in metropolitan cities. However, to generate an exact model for minimizing the health effects caused by air pollution, the prediction of fine-grained particles in the air is crucial. Due to globalization and industrialization, people tend to move from rural areas to cities. This rising population in the cities is the main reason behind the air pollution in cities. The continuous intake of polluted air may lead to severe health effects on people. Elderly people with heart, lung, and chronic diseases and children are more prone to breathing issues caused by the continuous intake of polluted air. So, it is essential to predict the quality of air in a region in order to prevent people from the harmful health effects of air pollution. Hence, an air quality prediction framework (AQPF) to assess the health effects caused by air pollution is generated in this work with the utilization of deep learning techniques. The deep-learning-based AQPF is developed to determine the concentration of air pollutants in the air in order to predict the health effects caused by them. The real-time data are used to create this model. The gathered real-time data are pre-processed. The pre-processed cleaned data are now considered for feature extraction. The statistical features, temporal features, and spatial features are all extracted from the cleaned data. The extracted features are now provided as the input to an optimal weighted prediction score-based ensemble learning network (OWPS-ELNet). The developed OWPS-ELNet is made of connecting the machine learning approaches like support vector regression (SVR), multi-layer preceptron neural network (MPNN), extreme learning model (ELM), bi-directional long short-term memory (Bi-LSTM), and recurrent neural network (RNN). The final classification scores are obtained from the OWPS-ELNet. The classification scores obtained ensemble models are given to the weighted prediction score fusion process. Here, the weights optimization for the weighted prediction score fusion is carried out with the help of the fitness-adapted reptile search algorithm (FA-RSA). From the final fused weighted prediction score, the quality of the ambient air is determined. From the determined air quality index (AQI), the health effects of the ambient air can be predicted accurately in that region. Extensive experiments are carried out with other previously generated AQPFs in order to prove the accurate prediction results and the efficient performance offered by the generated AQPF.