Air quality prediction based on residual Gaussian variational autoencoder
To enhance the accuracy of multivariate air pollution prediction,a Residual Gaussian-mixture Nested Factorial Variational Autoencoder(RNF-VAE)model was proposed.The model was evaluated using air pollution data from six monitoring stations in Huainan,Anhui,with performance assessed through three statistical metrics.The methodology involved constructing the RNF-VAE to handle multivariate air pollution data and comparing its performance with mainstream models,including Long Short-Term Memory(LSTM),Gated Recurrent Unit(GRU),Bidirectional LSTM(BiLSTM),and Bidirectional GRU(BiGRU).Results demonstrated that the RNF-VAE performed excellently in predicting six pollutants,achieving 35.12％reduction in RMSE,29.12％decrease in MAE,and 11.17％increase in R2.These findings indicated that the RNF-VAE provided higher accuracy and reliability in air pollution prediction,effectively addressing the complexity and data uncertainty in multivariate air pollution,thus offering valuable insights for policy formulation and environmental management.
air pollutantdeep learningresidual learningvariational autoencodermachine learning
万方数据
维普
