基于灰理论的中国股票市场短期组合预测建模研究

发布时间：2018-05-17 22:12

本文选题：灰理论 + 支持向量机　；参考：《武汉理工大学》2012年硕士论文

【摘要】：随着社会经济的发展和人民收入水平的提高,股票已经成为人们投资理财的一种重要工具。我国的证券市场目前还处于发展的初始阶段,其波动性和风险性都远远高于国外的成熟市场,因此准确地预测股价对于投资决策具有十分重要的指导意义。本文在灰色预测模型研究的基础上,将结合支持向量机的理论与方法构建组合模型对股票价格进行短期预测建模研究。主要内容如下：第一章主要介绍股票预测方法,概述了证券投资分析方法、数理统计方法、现代技术分析方法以及灰色系统理论等方法。第二章首先论述了中国股市不符合随机游走模型,股票价格的波动存在规律性,然后说明中国股市并没有达到弱式有效,从而说明中国股市在一定程度上是可以预测的。第三章首先介绍了灰色系统理论,接着针对GM(1,1)模型的模拟序列未能较好的反映出原始数据序列的光滑比和级比动态变化的问题,提出了基于光滑比和级比序列的GM(1,1)组合预测模型,并通过实证表明模型的有效性。然后针对股市存在涨跌停盘或短期节假日的情况,尝试将非等间距GM(1,1)模型运用到预测中,提出了通过累加法将灰导数优化和背景值优化进行组合,再采用逐步迭代来估计模型参数的新方法,实证表明该方法得到的模拟和预测值具有较高的精度。最后分析了经典GM(1,N)模型的建模机理,并阐述了经典GM(1,N)模型存在三个方面不足,并针对各个不足进行了相应改进,并提出了改进后的GM(1,N)模型,通过实证分析表明模型的有效性,能够运用到股价的短期预测中去。第四章将支持向量机解决小样本、非线性及高维模式识别的优势与灰色预测模型“小样本、贫信息”的特点相结合进行组合建模以及通过累加生成挖掘原始数据序列中潜藏的内在规律的特征相结合,提出了基于SVM的GM(1,1)模型的股价预测方法和基于SVM的GM(1,N)模型的股价预测方法,并且根据灰色理论中光滑比和级比的定义,提出了基于SVM的级比非线性灰色模型和基于SVM的灰色光滑比和级比预测模型,通过实证表明两者结合能够很好的运用到股价预测中,并且精度较高。第五章介绍了本文的主要研究内容、研究成果和创新点,并对未来的研究工作进行了展望。
[Abstract]:With the development of social economy and the improvement of people's income level, stock has become an important tool for people's investment and financial management. The stock market of our country is still in the initial stage of development, its volatility and risk are far higher than the mature markets abroad, so it is very important to predict the stock price accurately for the investment decision. In this paper, based on the research of grey forecasting model, combined with the theory and method of support vector machine (SVM), a combination model is constructed to study the short-term forecasting model of stock price. The main contents are as follows: The first chapter mainly introduces the stock forecasting method, summarizes the stock investment analysis method, the mathematical statistics method, the modern technical analysis method and the grey system theory and so on. The second chapter first discusses that the Chinese stock market does not conform to the random walk model, and the fluctuation of stock price is regular, and then shows that the Chinese stock market does not achieve weak efficiency, which shows that the Chinese stock market can be predicted to a certain extent. In the third chapter, the grey system theory is introduced first, and then the simulation sequence of the GM-1) model can not reflect the smooth ratio and the dynamic change of the order ratio of the original data sequence. A combined prediction model based on smooth ratio and order ratio sequence is proposed, and the validity of the model is demonstrated by empirical results. Then, aiming at the situation of stock market with fluctuation limit or short-term holiday, this paper tries to apply the non-equal-spacing GMM1Q1) model to the prediction, and puts forward the combination of grey derivative optimization and background value optimization by the accumulative method. A new method of estimating the model parameters by step-by-step iteration is adopted. It is proved that the simulation and prediction values obtained by this method have high accuracy. At last, the paper analyzes the modeling mechanism of the classical GM1N) model, and expounds the shortcomings of the classical GM1N) model in three aspects, and makes corresponding improvements to each deficiency, and puts forward the improved GM1N) model. The validity of the model is proved by the empirical analysis. Can be applied to the stock price forecast in the short term. In chapter 4, support vector machine is used to solve the advantages of small sample, nonlinear and high dimensional pattern recognition and grey prediction model. The characteristics of "poor information" are combined to combine the characteristics of the combination modeling and the cumulative generation of the inherent laws hidden in the mining original data sequence, In this paper, the stock price forecasting method based on SVM model and SVM model is put forward, and the definition of smooth ratio and grade ratio in grey theory is given. The nonlinear grey model based on SVM and the prediction model of grey smooth ratio and grade ratio based on SVM are proposed. The empirical results show that the combination of the two models can be applied to the stock price forecasting well, and the accuracy is high. The fifth chapter introduces the main research contents, research results and innovation points, and prospects for future research work.
【学位授予单位】：武汉理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F224;F832.51

【参考文献】