Short-term PM2.5 forecasting based on CEEMD-RF in five cities of China.

The development of industrial civilization has greatly enriched the material and spiritual life of human beings, but it is accompanied by the intensification of the consumption of earth resources and environmental pollution. The smog that has emerged in various parts of China in recent years is a typical problem, which not only endangers human health but also affects normal human work and life. It is difficult to control smog in a short time productively, so people need to understand the rule of smog formation gradually, and effectively predict the PM2.5 index to help people continuously analyze relevant mechanisms and timely protect-related hazards. This paper proposes a hybrid model that uses the Complementary Ensemble Empirical Modal Decomposition algorithm to mine the information in the original PM2.5 sequence and then predicts the pertinent random forests. The trend of PM2.5 concentration during the decomposition process is effectively reflected, and the decomposition sequence is modeled by the high tolerance of the random forest to the noise data and the good fitting ability. In the modeling process, the parameters are optimized according to the evaluation function of the model on the verification set, and eventually, the prediction sequences are superimposed to obtain the final predicted PM2.5 concentration value. The validity of the model is verified by the data of several Chinese cities with different geographical features in the past 5 years. The results show that the recommendation model is higher than other comparison models in terms of model stability and prediction accuracy.