基于PCA的相似度方法在半导体产品分类中的应用

Application of Similarity Based on PCA to Clssification of Semiconductor Productor

以传统的主元分析方法(Principal Component Analysis,PCA)为基础,利用负载矩阵构造SPCA相似因子(PCA Similarity Factor)和改进的相似因子SλPCA,通过衡量负载向量之间角度的大小对不同工况批次数据进行相似度比较,实现不同工况产品的分类.采用累计方差贡献率方法确定主元个数,将SλPCA应用到半导体间歇过程中进行数据分类.实验仿真结果表明:利用改进的相似因子SλPCA对半导体多工况数据的分类具有显著效果.

Based on the traditional principal component analysis （PCA）, the load matrix is used to get S_PCA （ PCA similarity factor） and S_PCA^λ. Similarity factors of data under different working conditions are got by measuring the size of the angle between loading vectors, thus, The productions obtained from different conditions can be classified. In this article, the number of principal components is got by cumulative percent variance,S_PCA^λ is applied to semiconductor process to classify the data. The simulation results show that S_PCA^λ performs well to the classify of semiconductor data from different operator conditions.