摘要
主成分分析是用于简化数据的一种技术,对于某些复杂数据就可应用主成分分析法对其进行简化。文中所用到的是一种连续统一的主分量分析法,它利用特征结构的正交性,提取出用于下一主分量的初始权向量,并且任何一种适用于线性前向反馈神经网络的主分量分析法都可作为此算法中的权修正等式。最后,将这种PCA法与普通PCA法运用于股票数据之中进行比较,结果对比证明用此方法提取出的数据比以前有所改进。
Principal component analysis is a technique used to reduce the complication of data .To some complicated data can use PCA to reduce them.The paper proposes a strictly local unified sequential method for principal component analysis.It uses orthogonality of the eigenstructure to extract the initial weight vector for the next component extraction. Any principal component analysis algorithm for linear feedforward neural networks can be used as the weight updating equation in our method. At last,compare it with another algorithm using the data of stock. The result shows that the unified sequential method is more effective.
出处
《微机发展》
2005年第2期67-68,72,共3页
Microcomputer Development
关键词
主成分分析
权修正等式
降级退化
principal component analysis
weight updating equation
degrade
