Classification of monofloral honeys based on their quality control data

Four hundred and sixty-nine samples of fir, cinder heather, chestnut, lavender, acacia, rape, and sunflower honey were characterized by their moisture, conductivity, diastase activity, pH, free acidity, color, hydroxymethylfurfural and percentage of fructose, glucose, saccharose, erlose, raffinose, and melezitose. A principal component analysis performed on the corresponding matrix yielded the formation of four clusters. A stepwise discriminant analysis allowed us to obtain 100% of good predictions with only conductivity, pH, free acidity and percentage of fructose, glucose, and raffinose as variables. The simulation performances of the model were estimated from an external testing set. (C) 2003 Elsevier Ltd. All rights reserved.