【摘要】 当神经网络用于上市公司的ST预测建模时,取得高质量的样本是相当重要的。本文连续运用主元分析(也称动态主元分析),将多年的数据应用到经济预测模型中去,既增大了信息量又没有增加网络的复杂性,使得预测更加合理有效。最后将动态主元分析与BP网络结合构造了一个网络模型,并给出了实证研究的详细结果。更多还原

【Abstract】 It is quite significant to get high-quality samples when neural network is applied to ST forecast model.We apply data of many years to ST forecast model by series principal component analysis dynamic principal component analysis.We therefore increase information quantity while do not increase network complexity.We propose a new ST forecast model by the combination of dynamic principal component analysis and back propagation neural network.We finally present some practical detailed results.更多还原