基于BP神经网络和GARCH模型的中国银行股票价格预测实证分析

发布时间：2018-02-13 20:36

本文关键词： BP神经网络 GARCH模型短期预测　出处：《兰州大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着中国金融市场与国际接轨,金融衍生品市场初步建成,金融投资工具在多样化、高杠杆的条件下也带来了巨大的金融风险.复杂多变的金融市场上对于金融投资分析工具的要求也就更高,催生出了多种对于股票价格预测的方法.对于不同的数据以及不同的市场环境需要不同分析方法.神经网络算法所具有的分布式存储数据以及学习反馈机制的特点使得它在预测等方面有独到的作用.本文中选取中国银行股票收盘价,采用BP神经网络(即前馈模型)和GARCH模型的方法对股票价格进行了预测,通过对比分析得出结论BP神经网络在隐含层节点数为5时对于市场数据拟合度最好；而GARCH模型在对股票价格预测方面也是有效的,主要是因为中国银行股票数据具有尖峰厚尾和平稳性特征.最终得出结论两种预测方法都能够对中国银行股票短期价格进行预测,但BP神经网络预测方法优于GARCH模型的预测方法.
[Abstract]:With China's financial market in line with international standards, the financial derivatives market has been initially established, and financial investment instruments are diversifying. Under the condition of high leverage, it also brings great financial risks. The requirements for financial investment analysis tools in complex and changeable financial markets are even higher. Different analysis methods are needed for different data and different market environment. The distributed storage data and learning feedback mechanism of neural network algorithm are special. In this paper, the closing price of Bank of China stock is selected. The method of BP neural network (i.e. feedforward model) and GARCH model are used to predict the stock price. Through comparative analysis, it is concluded that BP neural network has the best fit for market data when the number of hidden layer nodes is 5:00. The GARCH model is also effective in forecasting the stock price, mainly because the bank of China stock data has the characteristics of peak, thick tail and stability. Finally, it is concluded that both of the two forecasting methods can predict the short-term price of Bank of China stock. But BP neural network is better than GARCH model.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP183;F832.51

【参考文献】