关键词: 商空间, 动态粒度, 小波神经网络, 空气质量, 预测

Abstract: A new method of air quality forecasting based on dynamic granular wavelet neural network is put forward by the combination of quotient space theory, wavelets theory and neural network theory. Different granula can be obtained by the granulating of original data domain using quotient space theory and the best one can be found by testing it in the practice. The best guanula is used as the input to wavelet neural network for air quality forecasting. By this means, the forecast accuracy can be improved after the problem solving space has been changed. Experimental result of air quality forecasting also shows that this method is more effective.

Key words: quotient space, dynamic granula, wavelet neural network, air quality, forecasting