Deep Learning–Based Prediction of Urban Air Quality Using Multisource Environmental Data

Abstract

Accurate and fast air quality forecast systems are necessary because urban air pollution is a serious threat to both environmental sustainability and public health. The intricate nonlinear and spatiotemporal interactions included in urban air quality data are frequently difficult for traditional analytical and machine learning algorithms to capture. Using multisource environmental data, this study suggests a deep learning-based methodology for forecasting metropolitan air quality. To improve prediction accuracy, the suggested method incorporates a variety of data sources, such as air pollution concentrations, meteorological variables, mobility indicators, and land-use features. The temporal relationships and spatial correlations between monitoring stations are modelled using a deep neural networks architecture. The suggested neural network model outperforms traditional machine learning techniques in forecasting important air quality indicators, according to experimental results. The model is developed and tested using practical urban datasets, alongside its performance is evaluated using conventional metrics like RMSE, MAE, and R2. The results demonstrate the efficacy of deep learning in conjunction with multisource data fusion for accurate urban air quality estimation, providing insightful information for sustainable urban planning and environmental management.