Procrustes cross-validation
Create a validation set in your browser
Procrustes cross-validation (PCV) is a novel approach for validation of regression or classification models. It generates a new dataset — PV-set, which can be used for validation of PCA/SIMCA/PCR/PLS and other models in the same way as with an independent validation set.
Use one of the two algorithms:
- PCVPCA Algorithm based on Principal Component Decomposition. Works best for validation or data augmentation of one class classifier or discrimination model (create PV-sets for each class separately).
- PCVPLS Algorithm based on Partial Least Squares. Works best for validation or data augmentation of regression or discrimination model (use dummy response variable).
You can read more about the method in this paper (open access). You can also get PCV code for R, Python, MATLAB, Javascript in GitHub repository of the project.