节点文献
股票智能预测决策研究及应用
The Study on Intelligent Forecasting and Decision of Stocks and Its Application
【作者】 郭刚;
【导师】 戴冠中;
【作者基本信息】 西北工业大学 , 控制理论与控制工程, 2000, 博士
【摘要】 本文针对股票市场这一非线性系统进行了智能预测、决策分析的理论和方法研究,得出了一些有用的结论,解决了一些问题;同时结合上述理论的应用,对已有的股票分析系统,给出了改进意见和设计框架,无论从理论方面还是应用方面都是有价值的:1.根据模糊集与随机集之间的关系——随机集上的落影函数等价于模糊集 上的隶属函数;从而,转化模糊集隶属函数的计算为随机集落影函数的 求取。按照随机集落影函数的物理含义,推出单因素、多因素落影预测 的估计公式。给出模糊集值统计下落影函数的最优无偏估计定义,为在 线修改模糊规则给出理论基础。2.股票的筛选问题一直是一个非常令人烦恼的问题,通过模糊数学理论,作 者给出了一种非常有效的股票筛选方法,可以结合多种股票属性进行综 合评判,从而选择出最适合投资者投资的股票。3.基于股票的K线运行形态,充分结合已有的股票技术分析和人工智能方 法,设计了混合认知和模糊预测系统,预测结果表明,对股票在一定时期 内的多步预测问题,该种模型具有极高的准确度。本系统适合进行一段 时期的预测,同时具有自修正能力。4.对混沌动力学理论进行研究和探索,依据确定性混沌理论,发现了混沌 非线性系统智能建模模型的变量选取规律,提出一种进行非线性系统建 模变量选取的方法,通过仿真例子和股票价格的预测应用表明,无论是 对于典型的混沌系统的时间序列建模,还是对具有混沌特性的经济系统 的时间序列建模,采用本方法选取建模变量所建模型的拟合精度最高, 预测和泛化能力最好。5.提出一种基于混沌理论进行股票价格多步预测的方法。只需要考虑系统 是否混沌,然后依据混沌理论给出了进行股票价格多步预测明确的最大 时间尺度。对于其它复杂非线性系统的多步预测同样具有指导意义。6.预测股票的走向,选择出准备投资的股票后,依然需要解决组合投资比 例的问题,作者结合Markowitz模型,并进行了一些解法上的改进,可以 使投资者在明确的风险收益条件下,获得最大的收益。7.对组合投资比例权数为负数的问题进行了探讨,给出了解决办法,并给出 西北工业大学博士论文 股票智能预测决策研究及应用 详细的计算方法,解决了组合投资中不能卖空的问题。8.设计了大量的股票分析软件和辅助分析工具,并通过对现有股票分析系 统的分析,提出了理论结合应用的框架和解决思路,设计了新的股票分 析预测决策系统,无论对股票投资者还是系统解诀方案提供商,以及证 券商都是有参考价值的。5
【Abstract】 This dissertation mainly studies intelligent forecasting and decision making of stock market. Some new results are presented, and to be used to solve some problems in stock forecasting and decision making. These new methods are very useful to the field of financial applications as follows: Fuzzy statistics is used to get the membership functions of Fuzzy reasoning system and regulate fuzzy rules on line for better forecasting results. How to select suited stock to invest is very important. I put forward a new method with fuzzy logic to solve this problem, it can select the best stock for traders. A hybrid intelligent cognition system is provided. It is a combination of neural network and fuzzy logic, and is applied in the analysis and prediction of the trend of the stock market with satisfactory results. Using chaotic theory to calculate the embed dimension and relational dimension of a chaotic series or a normal economical series, we select the same number of variable to model and forecast this system. Examples show that, with this number of variable to model and forecast the modeled nonlinear-system, the forecasting error is minimal. Propose a new method for stock market forecasting. The chaotic theory is used to discover chaos in stock price time series. By calculating its Lyapunov exponent we found the max steps to predict the price of stock market. This method could also be used to other nonlinear system forecasting. According to Markowitz model, We introduce how to estimate the profit expectation and risk of portfolio by time series. And list two ways to solve portfolio selection model-Lagrange multiplication and critical line. A new simple method is given to solve the oversell problem of portfolio selection. After introducing the author抯 stock trading software , a new application architecture based on internet and data warehouse is presented. This software will benefit to traders and software developer.
【Key words】 intelligent forecasting; decision analysis; chaotic theoryt time-seriesforecasting; stock; portfolio selection; fuzzy logic; neural networks; fuzzyreasoniny; fuzzy statistics; nonlinear modeling;