Systems classification of air pollutants using Adam optimized CNN with XGBoost feature selection

Abstract

Since air pollutants released by motor vehicles have a bigger impact on human health than other air pollutants, air pollution is currently a very severe issue. Forecasting air quality has been used to control air pollution, as has public warning. The approaches existing today have many drawbacks like lower accuracy, lower performance, and high dimensionality problem, and hence to overcome all these drawbacks, the proposed method has been introduced. To classify the air pollutants in the dataset, a Convolution Neural Network (CNN), and eXtreme Gradient Boosting (XGBoost) based model (XGB-CNN) has been proposed. By using XGBoost to choose features from the pre-processed data, the number of parameters, high dimensionality issue, and training time are all decreased. Also, Adam optimizer is enhanced using power exponent learning rate to eliminate issues in fixed learning rate. CNN is used to categorize air pollutants into four types: sulfur dioxide (SO₂), ozone (O₃), nitrogen dioxide (NO₂), and carbon monoxide (CO) based on the Air Quality Index level of specified features. In terms of accuracy, F1-score, precision, recall and the effectiveness of the proposed XGB-CNN is compared to Decision Tree, Logistic Regression, k-Nearest-Neighbor, and Support Vector Machine, Long Short-term memory. It has also been shown that the proposed XGB-CNN method outperforms the existing systems in terms of efficiency by 5%.