基于小波去噪的我国股市分形分析

发布时间：2018-03-09 02:38

本文选题：分形市场理论　切入点：股市噪声　出处：《兰州商学院》2014年硕士论文　论文类型：学位论文

【摘要】：本文在分形理论和噪声理论的基础上，选取对我国股市具有代表性的上证综合指数，以股市重大事件为分界点，将2000-2013年划分为4个阶段，分析股权分置改革前后、金融危机期间和金融危机后噪声大小和分形度量，在此基础上，运用小波去噪对不同时期的收益率进行去噪，分析去噪前后噪声的变化及分形度量的变化，并用去噪后的股指价格序列进行预测，得到以下结论： 1.本文认为证券市场重大事件、投资者的非理性行为和信息不对称是造成股市噪声主要原因，运用方差比检验、Hurst指数和非对称性模型对不同时期的噪声大小作出一定的检验，发现我国股市是的分形市场，而且噪声较大，不同时期噪声大小明显不同，金融危机期间最大，股改前和危机后居中，股改后最小。 2.采用db2和coif小波函数对不同时期收益率进行2-4层的小波去噪，发现db2小波4层分解去噪效果最好；分析了db2小波去噪前后收益率分形度量和噪声比较，我们发现，去噪后R/S和ARFIMA计算的Hurst变大，而且更加显著，，这使得分形序列的长记忆性增强，具有较好可预测性。 3.利用2010-2013年经db2-4去噪后的股指收盘价进行EGARCH(2，2)建模并对40期价格进行预测，发现模型能较好的拟合价格走势，短期内有较好的预测效果。
[Abstract]:On the basis of fractal theory and noise theory, this paper selects the Shanghai Composite Index, which is representative of China's stock market, and divides the period of 2000-2013 into four stages, taking the major events of the stock market as the dividing point, and analyzes before and after the reform of the split share structure. On the basis of noise magnitude and fractal measurement during and after financial crisis, wavelet denoising is used to Denoise the rate of return in different periods, and the changes of noise and fractal measurement before and after denoising are analyzed. Using the de-noised stock index price sequence to predict, the following conclusions are obtained:. 1. This paper holds that the major events in the stock market, the irrational behavior of investors and the asymmetry of information are the main causes of stock market noise. The Hurst exponent and asymmetry model are used to test the noise in different periods. It is found that China's stock market is a fractal market, and the noise is large, the noise is obviously different in different periods, the biggest is during the financial crisis, the middle is before and after the stock reform, and the smallest is after the stock reform. 2. Using db2 and coif wavelet function to do 2-4 wavelet denoising in different periods, it is found that db2 wavelet decomposition has the best effect on denoising. After analyzing the fractal measure of yield and noise comparison before and after db2 wavelet denoising, we find that, After denoising, the Hurst calculated by R / S and ARFIMA becomes larger and more significant, which enhances the long memory of fractal sequence and has good predictability. 3.Using the stock index closing price after db2-4 de-noising from 2010-2013 to carry out EGARCH2 / 2) modeling and forecasting the 40-period price, it is found that the model can fit the price trend well and has a better forecast effect in the short term.
【学位授予单位】：兰州商学院
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51;F224

【参考文献】