RESEARCH ON WHITE AND RED WINE BLENDING IN THE PRODUCTION OF ROSE WINES BY MEANS OF THE PARTIAL LEAST-SQUARES METHOD

The difficult problem of recognizing wine blends in rose wines is assessed using a powerful statistical method called partial least squares. Genuine rose wines are first detected using a 280 nm/320 nm spectrophometric ratio. Blends of white and red wine are used as a calibration set and for an evaluation set of known composition. The proposed method evaluates the respective percentage of red and white wine in a blend and the error associated with the prediction. A comparison is made with principal component regression.