A combined forecasting structure based on the L1 norm: Application to the air quality. (15th September 2019)

• Record Type: Journal Article
• Authors: Wang, Biao; Jiang, Qichuan; Jiang, Ping
• Is Part Of: Journal of environmental management. Volume 246(2019)
• Page Start: 299
• Page End: 313
• Publication Date: 2019-09-15
• Subjects: Combined forecasting structure -- Phase space reconstruction -- L1 norm -- Multi-tracker optimization -- Sparse regularization; Environmental policy -- Periodicals; Environmental management -- Periodicals; Environment -- Periodicals; Ecology -- Periodicals; 363.705
• DOI: 10.1016/j.jenvman.2019.05.124
• Languages: English
• ISSNs: 0301-4797

Abstract

Abstract: Air pollution is very harmful to the industrial production and public health. Therefore, it is necessary to predict the air pollution and release air quality levels to provide guidance for public production and life. In most previous studies, pollutant data were directly used for predictions, which are rarely based on the structural characteristics of the data itself. Therefore, a novel combined forecasting structure based on the L1 norm was designed, aiming at pollution contaminant monitoring and analysis. It comprises analysis, forecast, and evaluation. Firstly, the original data are decomposed into several components. Subsequently, each component is expanded into a matrix time series by phase space reconstruction. The forecast module is then used to carry out the weighted combination of the prediction results of the three models based on the L1 norm to determine the final prediction result and the process parameters are optimized using the multi-tracker optimization algorithm. Moreover, comprehensive fuzzy evaluation was applied to qualitatively analyze the air quality. The daily pollution sources in three cities in China are taken as examples to verify the effectiveness and efficiency of the established combined forecasting structure. The results show that the architecture has a great application potential in the field of air quality prediction. Highlights: A novel combined structure consisting of analysis, prediction and evaluation. The model input structure is reasonably determined by phase space reconstruction. A L1 norm combination model based on multi tracker optimization is established. Comparative experiments are performed to prove the validity of the structure. The structure's applicability and effectiveness are verified in air pollution.