Predicting short-term PM2.5 concentrations at fine temporal resolutions using a multi-branch temporal graph convolutional neural network

The compressed package (study code.zip) contains the code files implemented by an under review paper ("Predicting short-term PM2.5 concentrations at fine temporal resolutions using a multi-branch temporal graph convolutional neural network").
<br>
Among the study code.zip, main.py is the model code based on a multi-branch temporal graph convolutional neural network. tgcn.py is the temporal graph convolutional network. utils.py contains some functions of graph convolution process. input_data.py is data processing.
<br>
The zip file (study data.zip) provides an example of air quality data including PM2.5 concentrations and some meteorological data. input_data.zip also contains a N by N adjacency matrix, which describes the spatial relationship between air quality monitoring stations.

该压缩包（study code.zip）内含一篇待审稿论文《基于多分支时间图卷积神经网络（multi-branch temporal graph convolutional neural network）的精细时间分辨率短时PM2.5浓度预测》所实现的代码文件。
在study code.zip中，main.py为基于多分支时间图卷积神经网络的模型代码；tgcn.py实现了时间图卷积网络（temporal graph convolutional network）；utils.py包含若干图卷积流程相关函数；input_data.py用于数据处理。
压缩包study data.zip提供了包含PM2.5浓度与部分气象数据的空气质量样例数据集；input_data.zip还包含一个N×N邻接矩阵，用于描述各空气质量监测站之间的空间关联关系。