Determining the number of principal components for best reconstruction

A well-defined unreconstructed variance (URV) is proposed to determine the number of principal components in a PCA model for best reconstruction. Unlike most other methods in the literature, this proposed URV method has a guaranteed minimum over the number pf PC's corresponding to the best reconstruction. Therefore, it avoids the arbitrariness of other methods with monotonic indices. The URV can also be used to remove variables that are little correlated with others and cannot be reliably reconstructed from the correlation-based PCA model. The effectiveness of this method is demonstrated with a simulated process. Copyright (C) 1998 IFAC.