论文部分内容阅读
本文通过分析某些实例,试图回答二个PCA应用中人们普遍遇到的问题,如何在实际问题中选择合适的分析方法以及如何选择合适的主成份。结论表明,当资料中各变量的重要性不取决于方差大小时以选择相关阵主成份法(R—PCA)为宜,重要性取决于方差大小时以选择协差阵主成份法(V—PCA)为宜。当PCA用于统计目的时,应通过对特征根进行显著性检验的方法确定选择多少个主成份,当PCA只是一种数据转换时,则无需作显著性检验,只需根据各主成份的人类学意义来确定如何加以选择。
By analyzing some examples, this article attempts to answer commonly encountered problems in two PCA applications, how to choose the appropriate analysis method in actual problems and how to choose the appropriate main component. The conclusion shows that when the importance of each variable in the data does not depend on the size of variance, it is appropriate to choose the relevant principal component method (R-PCA), the importance depends on the variance when choosing the covariance matrix principal component method (V- PCA) is appropriate. When the PCA is used for statistical purposes, the number of principal components to be selected should be determined by the significance test of the eigenvalue. When the PCA is only one type of data conversion, there is no need to make a significant test. The PCA The significance of learning to determine how to choose.