Research and application of a novel hybrid decomposition-ensemble learning paradigm with error correction for daily PM10 forecasting.

In this paper, a hybrid decomposition-ensemble learning paradigm combining error correction is proposed for improving the forecast accuracy of daily PM 10 concentration. The proposed learning paradigm is consisted of the following two sub-models: (1) PM 10 concentration forecasting model; (2) error correction model. In the proposed model, fast ensemble empirical mode decomposition (FEEMD) and variational mode decomposition (VMD) are applied to disassemble original PM 10 concentration series and error sequence, respectively. The extreme learning machine (ELM) model optimized by cuckoo search (CS) algorithm is utilized to forecast the components generated by FEEMD and VMD. In order to prove the effectiveness and accuracy of the proposed model, two real-world PM 10 concentration series respectively collected from Beijing and Harbin located in China are adopted to conduct the empirical study. The results show that the proposed model performs remarkably better than all other considered models without error correction, which indicates the superior performance of the proposed model. [ABSTRACT FROM AUTHOR]