摘要
根据动力学重建理论和多分辨分析的基本思想 ,利用小波变换重构股市系统的光滑吸引子 ,从而避开了预测的不适定问题 .以重构的状态矢量作为神经网络的多维输入 ,以上海证券交易所的上证指数为例 ,分别对 1 999年的 5 .1 9行情以及 2 0 0 0年的 2 .1 4行情后的几个关键点位进行预测 ,结果表明 。
According to the fundamental concepts of dynamical reconstruction theory and multiresolution analysis, a smooth attractor of the dynamical system of stock market is first unfolded in the reconstructed phase space in terms of wavelet transform. In this way the ill\|posed problem of prediction is solved. Then take the reconstructed state vectors as multiple inputs to neural network to predict stock index trend of Shanghai Stock Exchange (SSE) in future multiple trading days. The prediction results for the strongest bull market in history started on 19 May 1999 and two important values after the new biggest rise on 14 February 2000 show that the performance of our method is satisfied.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
2001年第8期19-23,共5页
Systems Engineering-Theory & Practice
基金
国家自然科学基金 ( 69872 0 30 )
国家教育部优秀青年教师基金 ( 1 997年度 )
陕西省自然科学基金( 98X0 8)
关键词
神经网络
多分辨分析
动力学重建理论
股票市场
证券交易
neural networks
multiresolution analysis
dynamical reconstruction
stock market trend prediction
