Open AccessThe Prediction of PM2.5 Concentration with an Intelligent Hybrid Model
Author(s) -
Xin Liu,
Xiaoxi Zhang,
Xuejing Zhao
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1624/2/022015
Subject(s) - support vector machine , hilbert–huang transform , artificial neural network , computer science , artificial intelligence , predictive modelling , residual , principal component analysis , data mining , machine learning , statistics , mathematics , energy (signal processing) , algorithm
The objective of this work was to propose a hybrid model to predict the concentration of PM2.5 in three cities of China. PM2.5 is one of the most important pollution worldwide, therefore effective prevention and control are beneficial to human’s production and life. A hybrid model, CEEMD-MFO-SVR-GRA-BPNN, was proposed to predict the concentration of PM2.5. The proposed model is the combination of (1) complementary ensemble empirical mode decomposition (CEEMD) to decompose the original PM2.5 concentration data; (2) support vector regression (SVR) to give a regressed prediction model in which parameters was optimized via moth-flame optimization algorithm (MFO); (3) grey relational analysis (GRA) to select atmospheric factors with distinguished effect on PM2.5; and (4) back propagation neural network (BPNN) to reduce the forecasting residual. The hybrid model was evaluated in three cities, Guiyang, Lijiang and Guangzhou of China, in which the environments and geographical locations are different. The implementation of the proposed model and well-known CEEMD-MFO-SVR, CEEMD- WOA-SVR, EEMD-MFO-SVR, EMD-MFO-SVR and MFO-SVR, BPNN-meteorology models, were compared. The results show that the prediction of the proposed hybrid model is more accurate than the compared models.