Based On The Wavelet Decomposition Of The Prediction Of Stock Prices And Extreme Low To High And Trading Decisions

The paper establishes a model to provide decision information for common investors reducing their cost in energy and getting large profit from price difference of the stock. The key point is to build a model to predict the period number of local maximum value of the highest stock price(simply called’extreme low point’) and local minimum value of the lowest stock price(simply called’extreme high point’) for a period of time in the future. For most people don’t have much time to observe the stock price trend every day and think about whether to buy in or to sell out, predicting the period number of the extreme high points of the maximum price and the extreme low points of the minimum price could be referenced by the investors. So they can observe stock market only at that time, and use some methods to predict the stock price to deal stocks and get relatively large price difference profit.The paper takes empirical analysis by data samples that the maximum and the minimum stock price of CSI300index Daily Line1000issue from September7,2007to October20,2011,and Shenzhen Development Bank Daily Line1152issue from August23,2006to October18,2011. First, briefly introducing wavelet analysis’s application in stock price prediction, and separately making wavelet decomposition to the time series of the highest price and the lowest price, extracting the low frequency parts. Then, make models to the low frequency parts of the time series of the highest price GA and the ones of the lowest price DA respectively, extracting the period number of extreme high points of the highest price GA and the period number of extreme low points of the lowest price DA, finding that both of them have a good linear relationship after mapping. Establish linear regression model to the two time series to predict the period number of the extreme points. Then match these extreme low points and extreme high points in time respectively to get groups of extreme high point and extreme low point, according to the results, making out the number of group to deal stocks. Finally, according to the number of group analyzed, building ARIMA model and linear regression model to stock time series of the lowest price at two numbers of periods before the lowest price appears, to predict price of the later two periods and compare the two models results, deciding the price to buy in. Use the same method to predict the price before the highest price come out in this group to decide the price to sell out. Buy low, sell high, get price difference profit.