Multifactor fuzzy time series forecasting modeling: an application in atmosphere pollutant concentration prediction

Effective prediction of atmosphere pollutant concentration is an essential and vital activity, for the benefit of conducting preventive measures to protect public health in advance. This significant activity has attracted extensive researches, of which the majority focus on addressing the prediction of specific value in the time series. Considering the imprecision of point prediction caused by the violent fluctuation in pollutant concentration, and the fact that the precise value is not as good as the degree of pollution adaptable for human perception and decision-making, a novel multifactor fuzzy time series (FTS) forecasting model for air quality level forecast is proposed in this paper. In this model, the fuzzy derivation mechanism of conventional FTS model is implemented by support vector machine (SVM), which enhances the systematization of the FTS model and overcomes its disability in simultaneously considering other influencing factors. Besides, to further improve the accuracy of our model, random forest (RF) and fuzzy information granulation (FIG) are employed to construct the optimal combination of influential variables. Finally, to verify the effectiveness of the proposed model, a case study and three kinds of contrastive models are performed, so as to further highlight the superior performance of the proposed model.