基于小波分解的股票价格极高点与极低点的预测及交易的决策

发布时间：2019-05-22 12:56

【摘要】：本文建立模型的目的在于为普通小股民提供决策信息,减少他们耗费在股票上的精力,并且能从股票差价中获得相对较大的利润。文章的关键点是建立模型预测未来一段时间股票最高价的局部极大值点(简称“极高点”)和最低价的局部极小值点(简称“极低点”)所处期数。由于大多数股民没有过多的精力,去每天观察股票价格走势,每天思考是否买进、是否卖出以及如何进行交易。那么预测最高价的极高点和最低价的极低点所处期数,就可以向股民提供参考,让他们在这段时间内观察股市。并且可以通过股价预测的方法给出买进和卖出的价格,从而进行交易,获得相对较大的差价利润。本文选取沪深300指数日线从2007年9月7日到2011年10月20日共1000期的最高价与最低价数据以及深发展日线从2006年8月23日到2011年10月18日共1152期的最高价与最低价数据为样本来进行实证分析。首先,简单介绍小波分析在股价预测上的应用,分别对最高价时间序列和最低价时间序列进行小波分解,提取其低频部分。然后,分别对最高价和最低价的低频部分时间序列GA.DA建立模型,提取出最高价GA的极高点所处期数以及最低价DA的极低点所处期数,对其作图发现他们分别具有良好的线性关系。对这两组时间序列建立线性回归模型,预测出极值点将会出现的期数。再次,这些极低点与极高点按时间先后分别配对,得到几组极高点与极低点。根据这个结果,观察分析得出应该进行交易的极低点和极高点的组数。最后,根据分析所得到的组数,在该组的最低价出现的前两期时,对股票最低价的价格时间序列建立ARIMA模型以及线性自回归模型,用这两个模型分别预测得出后两期的价格并将两个模型所得的结果进行对比,从而决定买进的价格。用同样的方法预测该组最高价出现之前的价格并进行对比,从而决定卖出的价格。这样先以低价买进,再以高价卖出,从中获取差价利润。
[Abstract]:The purpose of this paper is to provide decision-making information for ordinary small shareholders, reduce the energy they spend on stocks, and make a relatively large profit from the stock price difference. The key point of this paper is to establish a model to predict the number of periods in which the local maximum point (referred to as "the highest point") and the lowest local minimum point (referred to as the "very low point") of the highest price of the stock in the future. Because most shareholders don't have too much energy to watch stock prices every day, think about whether to buy, sell, and how to trade every day. Then predicting the number of periods between the highest peak and the lowest extreme low can provide a reference for shareholders to observe the stock market during this period. And the price of buy and sell can be given by the method of stock price prediction, so that the transaction can be carried out and a relatively large margin of difference can be obtained. This paper selects the data of the highest price and the lowest price of the 1000 issue of the Shanghai and Shenzhen 300 Index from September 7, 2007 to October 20, 2011, as well as the 1152 issues of the Shenzhen Development Japan from August 23, 2006 to October 18, 2011. The data are samples for empirical analysis. Firstly, the application of wavelet analysis in stock price prediction is briefly introduced. The highest price time series and the lowest price time series are decomposed by wavelet, and their low frequency parts are extracted. Then, the models of low frequency partial time series GA.DA with the highest price and the lowest price are established respectively, and the number of periods at the highest point of the highest price GA and the number of periods at the lowest point of the lowest DA are extracted. It is found that they have a good linear relationship. A linear regression model is established for these two groups of time series to predict the number of periods in which extreme points will appear. Thirdly, these extreme lows and extreme high points are matched according to the time order, and several groups of extreme high points and extremely low points are obtained. Based on this result, the observation and analysis shows the number of extremely low and extremely high points that should be traded. Finally, according to the number of groups obtained from the analysis, the ARIMA model and linear autoregression model are established for the lowest price time series of stocks when the lowest price of the group appears in the first two periods. The prices of the last two periods are predicted by the two models and the results of the two models are compared to determine the purchase price. The same method is used to predict and compare the prices before the highest prices in the group to determine the prices to be sold. In this way, buy at a low price, and then sell at a high price to get profit from the price difference.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;F224

【相似文献】