期刊
ATMOSPHERIC ENVIRONMENT
卷 237, 期 -, 页码 -出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.atmosenv.2020.117411
关键词
Air quality prediction; Social media; LSTM; 2015 China Victory day parade
资金
- National Key Research and Development Projects [2018YFB0505300, 2017YFB0503703]
- Guangxi Science and Technology Major Project [AA18118025]
- National Defense Science and Technology Innovation Special Zone Project
Air pollution, such as PM2.5 (particulate matter with an aerodynamic equivalent diameter of less than 2.5 mu m), PM10 (particulate matter with an aerodynamic equivalent diameter of less than 10 mu m), NOx, and SOx, is a global concern because it may cause many chronic and fatal diseases, especially in developing countries. To better address air pollution problems, an important step is the timely and accurate prediction of air quality. Traditional methods are mainly based on meteorological data, regression model data, remote sensing data and different retrieval methods. Numerous studies on deep learning methods have suggested that these approaches may be able to perform accurate predictions for complex systems. In this paper, a long short-term memory (LSTM) approach for predicting air quality is proposed; moreover, meteorological data are used and Chinese social media is investigated as a proxy for public perceptions and responses for air quality prediction. We gathered daily air quality data, meteorological data and Weibo check-in data for Beijing, China from January 1, 2015 to December 31, 2016. The average sentiment of the related Weibo posts was selected as the public response proxy. The performance of our proposed model is evaluated based on real data. The root-mean-square error (RMSE) and the mean absolute error (MAE) indicated that our method presented better prediction results than traditional methods in terms of the PM2.5, PM10, O-3, NO2, SO2 and CO concentrations. We focused on the prediction performance during the 2015 China Victory Day Parade period, during which social and political factors played an important role in air quality predictions. The results indicated that the proposed method, which incorporates public response data, was especially suitable for predicting the air quality in extreme short-term social events and provides a timely social measurement and feedback for environmental problems.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据