The Application of Set Pair Analysis on City Air Pollution Index Forecasting

The predictors of city air pollution forecast models which are specially selected have prefer forecast ability in general. But sometimes when the situation changes, some predictors are negative and may result in failure of the forecast model. According to the principle of Set Pair Analysis (SPA), regarding uncertainty and certainty as a dynamic systematic procession, the evolution of predictors' action in every forecasting is analyzed and processed dynamically. That is to say, the judgment of potential states and Identical-Discrepancy-Contrary Analysis are made about predictors before being used to calculate weather forecast, then the effect of predictors on weak potential state which may interfere forecasting is suppressed effectively, while those predictors on strong potential state which may be contributed to forecasting is given full play. As a result the dynamic evolution in the structure of predictors is made in the forecast model, and the rationality of forecasting mechanisms and the ability of models are intensified. So adding the processing uncertainty to the forecasting model is help to improve the forecasting accuracy.