Research On Wavelet Analysis And Neural Network Application In Stock Index

With the development of stock market and economics, more and more people invest the stock. Accordingly, prediction the trend of stock market is of great theoretical significance and applied value. As an important figure, the stock index is strongly uncertainty and non-linear. So the prediction always uses BP Neural Network, which is not easy. For the limitation of the BP Neural Network, the main idea of the dissertation shows that RBF is more suitable to prediction the stock index. At the same time, the dissertation uses the wavelet analysis theory, which makes a better prediction.The dissertation analyses the situation of China's stock market at first. Then, it introduces the wavelet analysis theory and describes the applied approach to prediction stock index. Next, the dissertation generalized the artificial neural network, illuminates the calculating method and regulation of BP neural network and RBF neural network. On the basic of theory, the dissertation chooses 452 closing index of China's security market as analysis samples from 1997 to 2006, combining the wavelet analysis and neural network with the application of prediction stock index.The fourth and fifth chapter of the dissertation is the empirical research. Chapter four of this article using wavelet deals with sample data, using sym8 wavelet function in signal denoising thus the data would be more smoothing and accurate of prediction would be improved. In the chapter five the article uses BP NN making model; takes the data after wavelet denoising as network inputting, and calculates MAE and MAPE of prediction result; takes the same inputting data using RBF NN making model. finally the article compares BPNN prediction result with RBFNN prediction result, It is conclusion that RBF NN is more accurate than BP NN in stock index prediction. In order to prove the effectiveness of wavelet theory's application in stock index prediction, this article makes a comparison between the predicted result without and with wavelet denoising. In that case wavelet theory is proved to be effective in application.