Hourly PM2.5concentration forecasting based on feature extraction andstacking-driven ensemble model for the winter of the Beijing-Tianjin-Hebeiarea

Özet

Precise forecasting of hourly PM2.5concentration is essential for its monitoring and controlling, especially for thewinter of the Beijing-Tianjin-Hebei area as severe haze episode occurs frequently. Therefore, this paper exploresan hourly PM2.5concentration forecasting technique based on the Stacking-driven ensemble model and twokinds of input selection methods. Firstly, the partial autocorrelation function (PACF) is employed to selectfeatures of time-lagged factors, meanwhile, Spearman correlation coefficient is introduced to extract latentfeatures of exogenous factors, jointly determining the input of the forecasting model. Subsequently, a two-layerStacking-driven ensemble model is formed as the prediction model to enhance the feature representation andinformation utilization capacities. Among this ensemble model, BPNN, IBPNN, and ELM are used as base-modelwhile LSSVR is the meta-model. Moreover, four-fold cross-validation is carried out in each base-models to en-hance the generalization performance of the model. Finally, the output of each base-model is new input for meta-model to acquire ultimate forecasting values. Case study of hourly PM2.5concentration forecasting in the Beijing-Tianjin-Hebei area during the winter substantiates that: (1) application of input selection methods is beneficial tosatisfactory forecasting results; (2) the forecasting performance of the Stacking-driven ensemble model is farbetter than any single models composing of it; (3) the proposed model can tackle highly complicated and ex-tremely high concentration PM2.5data, which is feasible to PM2.5concentration forecasting during the winter;(4) the proposed model with small forecasting error, strong generalization performance, as well robust fore-casting ability is potential in the early air warning systems.