小波信号消噪在趋势交易中的应用研究

发布时间：2018-03-05 16:07

本文选题：小波分析　切入点：消噪　出处：《上海交通大学》2012年硕士论文　论文类型：学位论文

【摘要】：本文从噪声交易着手，引入对于市场有效性的质疑，介绍了挑战有效市场假说的三种理论，分别是市场微观结构理论，，行为金融学和分形市场假说。基于以上介绍，本文认为市场是不完全有效的，而正是基于此，趋势交易才能长期存在于市场。然而，趋势交易也存在其特有的问题，那就是在震荡行情下其收益率回撤可能会过大，以至于引发可能的清盘风险。为了减小可能的收益率回撤，本文试图通过小波消噪将行情中的短期的交易噪音过滤掉，突出行情的趋势性，减少噪声交易对趋势交易产生干扰，从而提高趋势交易的绩效。为此，本文着重介绍了小波分析的相关理论，并以此为基础，应用HAAR小波对股指日线数据，股指15分钟数据，上证综指，深证成指，深ETF100日线数据等原始行情进行了一级分解和二级分解，并对各级分解的小波系数进行消噪，使用的消噪方法有默认阀值小波消噪方法，给定阀值小波消噪方法和强制小波消噪方法，最后得到不同的消噪后的行情。通过比较同一套趋势交易策略在原始行情及其相应的消噪后行情上的历史回测绩效，验证小波消噪对于提升趋势交易绩效的作用。实证结果表明小波消噪对于趋势交易绩效确实有显著的提升作用。本文还发现，小波二级分解效果普遍优于一级分解，最简单的强制小波消噪方法效果并不比其他算法相对较为复杂的小波消噪方法差。基于以上实证结果和发现，本文认为小波消噪技术在趋势交易中具备相当的应用价值。本文涉及的小波变换及消噪在MatLab上编程完成，趋势交易策略代码编写及历史绩效回测在行情分析系统MultiCharts上完成，相关代码参见附录。
[Abstract]:This paper starts with noise trading, introduces the doubts about the effectiveness of the market, and introduces three theories that challenge the effective market hypothesis, which are market microstructure theory, behavioral finance and fractal market hypothesis respectively.
Based on the above introduction, this paper thinks that the market is not fully effective, which is based on this, the long-term trend trading can exist in the market. However, the trend of trading still has its problems, it is in the market shocks the yield retracement may be too large, that may lead to the risk of liquidation.
In order to reduce the possible return rate, we try to filter out the short-term trading noise through the wavelet denoising, highlighting the trend of the market, reducing the interference of the noise trading to the trend trading, and improving the performance of the trend trading.
Therefore, this paper introduces the theory of wavelet analysis, and on this basis, the application of HAAR wavelet on stock index daily data, the stock index data of 15 minutes, Shanghai, Shenzhen, deep ETF100 data of the original daily market level decomposition and two level decomposition, and the decomposition levels of wavelet coefficient denoising. Denoising method using the default threshold wavelet denoising method, wavelet threshold denoising method is given and the forced wavelet denoising method, finally obtains the different denoising after market.
By comparing the same trend trading strategy in the original market and the corresponding market after denoising the history test performance, verify the wavelet denoising for improving the performance of trend trading role. The empirical results show that the wavelet denoising has significantly improved function for the performance of trend trading.
It is also found that the effect of wavelet two level decomposition is generally better than that of the first order decomposition, and the simplest coercive wavelet denoising method is no more effective than other algorithms.
Based on the above empirical results and discovery, this paper holds that the wavelet denoising technology has considerable application value in the trend transaction.
The wavelet transform and de noising in MatLab are programmed. The trend trading strategy code compilation and historical performance re test are completed on the market analysis system MultiCharts. The related codes are attached to the appendix.

【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;F224

【参考文献】