Dynamic graph convolution neural network based on spatial-temporal correlation for air quality prediction

Deep-learning architecture for PM2.5 concentration prediction: A review

“… 2022 New Delhi, India HGNN H/S/T+8 19.83 16.61 - - Dun et al. [ 30 ] 2022 Beijing/Fushun, China DGC-MTCN H/S/T+1 9.77/12.96 5.54/8.39 - 0.95/0.91 …”

“…Graph convolutional neural network (GCNN)-based methods . In previous studies [ [26] , [27] , [28] , [29] , [30] , [31] ], various studies have proposed GCNN models for PM 2 . 5 prediction.…”

“…Dun et al. [ 30 ] proposed a dynamic DGCNN that can update the graph structure based on spatial-temporal correlations among air pollutant concentrations. The proposed DGCNN can effectively capture both spatial and temporal dependencies and dynamically update the graph structure to capture changes in spatial-temporal correlations over time.…”

A novel encoder-decoder model based on Autoformer for air quality index prediction

“…(2) Many methods are available to improve the performance and efficiency of deep learning-based predictive models. These methods mainly include data preprocessing [20], deep learning model improvement [32,33,35], improvement of neural networks based on optimization algorithms [38,39], and other hybrid models [25,26,28]. We continue to experiment with various methods to improve model prediction accuracy and efficiency, such as feature selection, multi-objective optimization techniques, model improvement, etc.…”

“…Liu et al [ 27 ] proposed an attention-based air quality predictor (AAQP) to forecast the air quality index of Beijing in the future. Dun et al [ 28 ] proposed a DGC-MTCN model, which combined dynamic graph convolutional network (DGC) and multi-channel temporal convolutional network (MTCN) to predict the PM 2.5 in Beijing and Fushun and achieved good prediction accuracy.…”