基于灰关联理论和神经网络的价值预测方法

为了预测股票价格的短期走势,在预测算法中引进RBF神经网络,利用RBF神经网络具有唯一最佳逼近、无局部极小、学习速度快的特点,在预测股票行情时,能达到较高的精度。同时,为了优化RBF网络的输入参数结构,引入二次参数的概念,设计了基于灰关联理论的技术指标选择控制器,从众多的技术指标中选出部分最能反映股票近期趋势的指标,从而获得包含股市本质信息的低维输入,大幅度减少了运算量。最后,在综合两者优势的基础上构造了一种新型价值预测系统,该系统具有较快的运算速度和较高的预测精度。仿真实验表明,该方案是可行的。

Value Prediction Methods Based on Grey Relation Theory and Neural Network

To predict the short-term tendency of stock-price,the Redial Basis Function(RBF) neural network is introduced to the forecast algorithm.And because RBF has many excellent characteristics for nonlinear prediction,such as:optimal approximation,non-local minimum and short learning times,the high precise result can be obtained when RBF is used to predict stock market.At the same time,in order to optimize the structure of input parameters of the RBF neural network,a notion of secondary parameter is introduced and a kind of qualification-selecting controller based on the grey relation theory is designed.The controller can select some indexes which reflect the recent trend of stocks greatly from numerous technical indicators to realize the aim of a few inputs including more essential stock-market information and a large margin reduction in operation amount.Finally,a novel system of value prediction is designed on the basis of synthesizing RBF neural network and the controller’s advantage,and it has the faster operation tempo and the higher forecast precision.The results of numerical simulations demonstrate that the system is effective.