基于门控循环单元和图神经网络的PM2.5预测

PM2.5 Prediction Based on Gated Recurrent Unit and Graph Neural Network

PM2.5浓度指数是衡量空气质量的重要指标之一,但由于PM2.5数据的非线性以及受多种气象因素的影响,因此实现精确预测较为困难。本文采用门控循环单元和图神经网络相结合的混合模型预测的方式,并进一步采用改进的门控循环单元提升网络效果。通过对门控循环单元的输入信息与隐藏层信息进行数据交互增强上下文信息联系,使得门控循环单元模块的转移函数依赖于信息上下文。一系列实验结果表明:提出的改进方法相比于现有的方法具有更好的性能以及泛化效果,在中国生态环境部提出的京津冀地区真实数据集上验证了方法的有效性,与现有网络相比预测准确率更高。

PM2.5 concentration index is one of the important indicators to measure air quality. However, it is difficult to accurately predict PM2.5 due to the non-linearity of PM2.5 data and the influence of various meteorological factors.In this paper, the hybrid model prediction method combining Gated Recurrent Unit and Graph Neural Networks is adopted, and the improved Gated Recurrent Unit is further adopted to improve the network performance. Through data interaction between input information of Gated Recurrent Unit and information of hidden layer, the connection of context information is improved, so that the transfer function of Gated Recurrent Unit module depends on the information context. A series of experimental results show that the proposed improved method has better performance and generalization effect than the existing method, and the validity of the proposed method is verified on the real data set of the Beijing-Tianjin-Hebei region proposed by the Ministry of Ecology and Environment of China, and the prediction accuracy is higher than that of the existing network.