A balanced social LSTM for PM2.5 concentration prediction based on local spatiotemporal correlation. (March 2022)

• Record Type: Journal Article
• Authors: Shi, Lukui; Zhang, Huizhen; Xu, Xia; Han, Ming; Zuo, Peiliang
• Is Part Of: Chemosphere. Volume 291:Part 3(2022)
• Publication Date: 2022-03
• Subjects: Long short-term memory -- PM2.5 concentration prediction -- Spatiotemporal correlation; Pollution -- Periodicals; Pollution -- Physiological effect -- Periodicals; Environmental sciences -- Periodicals; Atmospheric chemistry -- Periodicals; 551.511
• DOI: 10.1016/j.chemosphere.2021.133124
• Languages: English
• ISSNs: 0045-6535

Abstract

Abstract: Reliable prediction for the concentration of PM2.5 has become a hot topic in pollution prevention. However, the prediction for PM2.5 concentration remains a challenge, one of the reasons is that current prediction methods do not consider the relevance of PM2.5 concentration among surrounding areas. In this paper, we propose the assumption that the PM2.5 concentration has spatial interaction, which includes two parts: 1) The PM2.5 concentrations observed by adjacent stations usually present relevant trends; 2) Stations with higher PM2.5 concentration tend to show higher influences on neighboring areas. Based on the spatial interaction assumption, we propose a balanced social long short-term memory (BS-LSTM) neural network for the prediction of PM2.5 concentration. BS-LSTM is composed of two kernel components: a social-LSTM based prediction model and a new balanced mean squared error (B-MSE) based loss function. On the one hand, to capture the spatiotemporal correlation of the PM2.5 concentration among adjacent stations, we develop a social-LSTM based model which has advantages in describing the trend information of neighboring locations. On the other hand, considering the unbalanced influence caused by various local pollution levels, we design a new B-MSE loss function to assign different attention to the observation stations. In the experiments, we evaluate the proposed method on two real-world PM2.5 datasets. The results indicate that BS-LSTM is promising, especially in the case of heavy pollution. Graphical abstract: Image 1 Highlights: A BS-LSTM model is proposed for PM2.5 concentration prediction. A social-pooling layer is utilized to capture the local spatial correlation. A new loss function to describe the influence of different local pollution levels. The validity of BS-LSTM is verified on two real-world PM2.5 datasets.