摘要
日益膨胀的股票市场信息远超出人们的处理能力，股票价格变得越来越难以预测。神经网络方法可以模拟人工智能处理海量信息。提高对股票市场的预测水平。运用中国1998-2005年股票市场数据，利用梯度下降法拟合了一个BP神经网络模型，在实证过程中重点讨论预测过程中出现的分类标准、过抽样、过度训练等问题。认为正确运用神经网络方法可以提高预测分析效果，神经网络模型可以谨慎地作为一种股票投资分析方法加以运用。
As the barometer of economics, the stock market is affected by many factors. Further, the growing information makes it impossible to forecast the stock price. This paper investigates the usage of the neural networks in stock investment and builds a BP neural network. In the process of empirical study, the author firstly focuses on the criterion of classification, over- sampling, over- training on building models; then the aothor presents some means to deal with the problems; finally the author comes to a conclusion： as a developing method, correct neural network can help investor beat the stock market.
出处
《统计与信息论坛》
CSSCI
2008年第1期63-67,共5页
Journal of Statistics and Information
关键词
BP神经网络
分类标准
过抽样
BP neural network
classification criterion
over- sampling
