基于全连接和LSTM神经网络的空气污染物预测

在空气污染日益严重的情况下进行空气污染物的预测工作是十分必要的。针对城市的空气污染物预测,提出了一种基于神经网络的混合模型方法:通过全连接神经网络方法,结合长短期记忆网络(Long Short-Term Memory,LSTM)方法,将历史空气污染物数据与大气数据进行空间与时间上的挖掘分析。运用全连接和LSTM两种神经网络方法混合的形式,与传统的单一模型方法相比,不仅能摆脱单一模型特征空间的局限性,还能提高预测的精度,具有更大的应用性和操作性。最后,以武汉市为例通过实验证明该混合模型较单一模型在空气污染物预测上具有更高的精度。

The Prediction of Air Pollutants Based on Full Connection and LSTM Neural Network

Air pollution has been an increasing challenge for many countries nowadays. It is of great necessary to forecast the quantity and spatial distribution of air pollutants. This paper proposes a novel ensemble method for air quality forecasting based on neural network. Via the combination of full-connection neural network and LSTM, the different spatial-temporal features of air pollutants concentration data and weather data are obtained with ensemble method. Comparing with traditional single methods, the ensemble method which depends on the form of ensemble of full-connection neural network and LSTM can not only get over the limitations of single model, but also improve the accuracy of forecasting. Finally, taking Wuhan as an example, the experimental results show that the hybrid model is more accurate than the single model in predicting air pollutants.