模糊神经网络在股票预测中的应用研究

发布时间：2018-03-12 15:10

本文选题：模糊逻辑　切入点：神经网络　出处：《安徽财经大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着我国经济体制改革和金融体制改革的深入,股票投资已经成为了社会生活的一个重要部分。股票价格的预测成为投资者关心和研究的重点。但股票市场是一个极其复杂的非线性动力学系统,具有高噪声、严重非线性和投资者的盲目任意性等因素以及各因素间的相关性错综复杂,造成其价格的波动往往表现出较强的非线性特征。另外,股市的建模和预测所处理的信息量往往是十分庞大的,对算法的要求很高,正是由于其复杂的非线性特征,使得关于股市预测的结果往往难如人意。如何建立一个运算速度和精确度都比较高的股市预测模型,对于金融投资者具有理论意义和实际应用价值。本文针对在股票价格预测中存在的困难引入模糊逻辑和神经网络的概念,利用模糊逻辑中可以用模糊性的自然语言表现知识和可以用Max, Min这类简单运算实现知识的模糊推理的特点,以及利用神经网络中能够生成不需要明确表现知识的规则和其强大的自学能力的特点,把二者结合起来构成模糊神经网络。利用TS模糊规则和前馈神经网络的方法进行建模,并且探讨了网络的结构、隐节点个数确定的原则、样本数据的选取和处理、初始参数的确定等问题。根据模糊神经网络对股票预测的原理,建立基于模糊神经网络的股市预测模型,并利用相关性分析对股票预测时的输入项进行了筛选。通过MATLAB7.0软件,对选取的绿景地产、潍柴动力、招商证券、宝钢股份和上证指数进行实证分析,根据神经网络常用的预测性能的评价指标对预测结果进行了评价,证实了该模糊神经网络进行预测是有效的,预测系统是成功的。
[Abstract]:With the deepening of China's economic and financial system reform, Stock investment has become an important part of social life. The prediction of stock price has become the focus of investors' attention and research. But the stock market is an extremely complex nonlinear dynamic system with high noise. Factors such as severe nonlinearity, blind arbitrariness of investors, and the correlation among various factors are complicated, resulting in their price fluctuations often showing strong nonlinear characteristics. In addition, Stock market modeling and forecasting often deal with a huge amount of information, and the algorithm is very demanding, precisely because of its complex nonlinear characteristics. How to establish a stock market forecasting model with high calculation speed and accuracy is of theoretical significance and practical application value for financial investors. In this paper, the concepts of fuzzy logic and neural network are introduced to solve the difficulties in stock price prediction. The characteristics of fuzzy logic can be used to express knowledge in natural language and simple operations such as Maxand Min can be used to realize fuzzy reasoning of knowledge. Using the characteristics of the neural network which can generate the rules which do not need to express the knowledge clearly and its powerful self-study ability, the fuzzy neural network is constructed by combining the two features. The model is modeled by using TS fuzzy rules and feedforward neural networks. The structure of the network, the principle of determining the number of hidden nodes, the selection and processing of sample data, and the determination of initial parameters are discussed. According to the principle of stock forecasting based on fuzzy neural network, the stock market forecasting model based on fuzzy neural network is established, and the input items in stock forecasting are screened by correlation analysis. The selected green scene real estate is selected by MATLAB7.0 software. Weichai Power, China Merchants Securities, Baosteel shares and Shanghai Stock Exchange Index are empirically analyzed, and the prediction results are evaluated according to the commonly used performance evaluation indexes of neural networks. It is proved that the fuzzy neural network is effective in forecasting. The prediction system is successful.
【学位授予单位】：安徽财经大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;TP18

【参考文献】