Partial least squares fine-tuning by a bootstrap estimated signal–noise relation to weight the loadings

This work analyses some partial least squares regression (PLSR) problems that can conceptually be improved. The result is the incorporation, in the classical PLSR algorithm, of a weighting stage giving rise to partial least squares with weighted loadings (WL-PLS) algorithm. Then some studies with real data have been done to evaluate its performance.