I am no data scientist, but reading on the subject it seems that PCA can be a bad idea in many cases and Regularization should be used instead - have you any experience with it?