摘要
本文首先论述了股市时间序列中的明显随机性 ,可能是由于非线性确定性系统中混沌行为的缘故 ,利用混沌的确定性可以进行短期预测。混沌时间序列预测首先要重构相空间 ,接着充分利用小波变换时频分析的局部化特性 ,提出了一种改进的小波网络结构 ,探讨了股市预测模型问题。经实例验证 ,该方法能有效地提高预测精度 ,避免了人工神经网络模型和指数自回归的固有缺陷。
This paper briefly discusses apparent randomness in stock market time serials first; it may be due to chaotic behavior of a nonlinear but deterministic system. It is possible to make short term prediction by using the determinism. This is done by making full use of the advantages of wavelet transform time frequency localization, the paper proposed an improved wavelet network structure, and the model for stock market prediction was considered. It can be seen from the example that this method can effectively improve the prediction accuracy, the intrinsic defects of artificial neural network and exponent auto regression are avoided.
出处
《管理工程学报》
CSSCI
2002年第2期32-37,共6页
Journal of Industrial Engineering and Engineering Management
基金
国家自然科学基金资助项目 ( 6 98740 0 4)
校高级人才资金资助 ( 16 830 0 0 0 2 7)
关键词
小波神经网络
股市预测
相空间重构
混沌
Wavelet neural network
stock market prediction
phase space reconstruction
chaos
