股票市场预测的组合方法研究

发布时间：2018-02-21 03:35

本文关键词： 股市预测灰色理论马尔柯夫过程 BP神经网络遗传算法组合模型　出处：《华中科技大学》2013年硕士论文　论文类型：学位论文

【摘要】：股票已成为金融市场中不可缺少的组成部分，股票价格跌宕起伏，股票市场风云变幻，不仅受内部规律的影响，而且还受到外部环境像政治、经济等因素的影响，但投资者为了获取丰厚的收益，还是希望能够准确分析和预测股票价格。那么如何建立一个比较理想的预测模型，是众多学者一直研究的内容。本文选取了两种组合模型分别对上证综合指数进行分析、预测和比较，两种组合方法分别为灰色-马尔柯夫模型和遗传算法优化的BP神经网络模型。首先，建立了灰色系统模型GM(1,1)，它所描述的是灰色变量，比较适合于数据少、短期的、波动小的系统对象，而对于波动性比较大的数列，拟合度比较差，预测精度不够理想。针对灰色模型这一缺点引入了马尔柯夫过程，然后运用组合方法灰色-马尔柯夫模型对上证指数进行预测；其次，通过对BP神经网络的建模与分析，发现它在局部极小点和收敛速度上存在一些问题，并引入了遗传算法的分析和研究，建立了第二种组合模型基于遗传算法的BP神经网络，并对单一的BP神经网络作了比较，得出组合模型预测效果更好的结论；最后，，通过两种组合方法的比较，来分析出它们各自所呈现的数据特征，文章的最后又对数据波动比较大的上证指数进行了预测。实验结果表明：两种组合方法在数据平稳和数据波动的情况下，从预测相对误差上来衡量，GM(1,N)-Markov模型整体上要比GA-BP模型的预测精度高些。
[Abstract]:The stock market has become an indispensable part of the financial market. The stock price has fluctuated and the stock market has changed. The stock market is not only affected by internal laws, but also by external factors such as politics and economy. However, in order to obtain rich profits, investors still hope to accurately analyze and predict the stock price, so how to establish an ideal forecasting model is the content that many scholars have been studying all the time. In this paper, two combination models are selected to analyze, predict and compare the composite index of Shanghai Stock Exchange. The two combination methods are gray Markov model and genetic algorithm optimized BP neural network model. In this paper, a grey system model, GM1 / 1, is established, which describes grey variable, which is suitable for the system object with less data, short term and low fluctuation, but the fitting degree is poor for the series with high volatility. Aiming at the shortcoming of grey model, the Markov process is introduced, and then the combination method is used to predict the index of Shanghai Stock Exchange. Secondly, through the modeling and analysis of BP neural network, It is found that there are some problems in the local minima and convergence rate, and the analysis and research of genetic algorithm are introduced. The second combinatorial model BP neural network based on genetic algorithm is established, and the single BP neural network is compared. Finally, through the comparison of the two combination methods to analyze their respective data characteristics, the last part of the paper forecasts the Shanghai Stock Exchange Index which has a large data fluctuation. The experimental results show that the prediction accuracy of the two combined methods is higher than that of the GA-BP model on the whole, when the data is stable and fluctuating.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51;N945.12

【参考文献】