Hybridized Deep Neural Networks for Air Pollution Forecasting and AQI Level Identification

Metropolitan air pollution is a serious problem across the world that is becoming worse because cities are growing quickly, the population is growing, and industrial activity is increasing. This extraordinary halt in residents' consumption and industrial output has effectively reduced air pollution emissions, creating typical and natural test sites to assess the impacts of human activity regulation on air pollution control and reduction. Air pollution are bad for people's health and the environment. Earlier studies have used machine learning (ML) and statistical modelling to classify and forecast air pollution. This work created a hybrid deep learning (HDL) system for predicting air pollution and a way to classify the air quality index (AQI) during the pandemic. At the beginning, the HDL model prepares the real air quality data for subsequent processing. The reported HDL model used an HDL-based air quality prediction and AQI classification framework, whereby the HDL was generated using a convolutional neural network employing an extreme learning machine (CNN) method. To best adjust the hyperparameter values of the HDL model, it is run. The results of the experiment show that the HDL model has a promising ability to classify predictions. We used the coefficient of determination R2, the mean absolute error (MAE), and the root mean squared error (RMSE) to test this model, which was made using the Python platform. The recommended HDL approach is better than well-known models like XGBoost, support vector machines (SVM), random forest (RF), and the ensemble model (EM). It has an R2 of 0.982, an RMSE of 16.522, and an MAE of 11.129.