A well-defined variance of reconstruction error (VRE) is proposed to determ
ine the number of principal components in a PCA model for best reconstructi
on. Unlike most other methods in the literature, this proposed VRE method h
as a guaranteed minimum over the number of PC's corresponding to the best r
econstruction. Therefore, it avoids the arbitrariness of other methods with
monotonic indices. The VRE can also be used to remove variables that ape l
ittle correlated with others and cannot be reliably reconstructed from the
correlation-based PCA model. The effectiveness of this method is demonstrat
ed with a simulated process. (C) 2000 IFAC. Published by Elsevier Science L
td. All rights reserved.