Research And Application Of Wavelet Neural Network In Stock Forecasting

The stock market has been in highly dynamic state.Sometimes,the share price will decrease a lot in one day,but it will occured that the share price would have been in risig trend for a long time.It seems that it is very difficult to predict the continuous changing stock market.In addition,people believe that the share price’s movement is a completely random movement pattern. The distribution of Stock returns appears to be in independent state,it seems that there is no link between share price.Usually we are often at great risk to do stock deals,for example, at one moment,you made much profit,but before long the share price suddenly dropped, you were likely to go bankrupt.Although there is a great risk to do stock deals, but it is undeniable that doing stock deals can really bring great benefits to people in a short time. If we can predict the stock price, which will bring participants who do stock deals much profits that you can not believe. So people’s enthusiasm for predicting stock price is unabated.Given the enthusiasm for researching about how to predict the stock price, so there has been many methods on stock prediction. Early time,people put linear methods in use, such as security investment analysis and time series analysis,but it is limited to predict the complex nolinear share system by using linear analysis method, and the effect of prediction is not undesirable. In recent years, people began to focus on the nonlinear prediction. For example, neural network, wavelet analysis, wavelet neural network predicting method that this article will introduce.Wavelet neural network combines the both advantages of wavelet analysis and neural network. Because wavelet’s participation,the performance of neural network is more stable, more powerful in function approximation ability. Thus the prediction effect of stock is better by using the method. But the wavelet neural network’s algorithm is BP algorithm, and BP algorithm is a local search algorithm, so it is easy to get the local solution. This paper presents to set a global search algorithm called genetic algorithm as algorithm of wavelet neural network.Then respectively to make use of the two different algorithms to train the wavelet neural network, then predicting the stock price by using the trained network and comparing their prediction results to prove the predicting effect of genetic algorithm is better than the BP algorithm.