According to the basic principle of pattern identification of neural networks, the paper proposed a classified wavelet network for forecasting stock market based on the property that the wavelet can smooth datum and adapt to the changes of function. Furthermore, according to the theory of ”pattern reappearance” of stock technical analysis and the characeristic that includes pattern memorizing and connecting, we establish stock market forecasting models for forecasting ”buying opportunity” of composite index based on classified wavelet networks.