Predicting the buckwheat flour ratio for commercial dried buckwheat noodles based on the fluorescence fingerprint

A rapid method for predicting the buckwheat flour ratio of dried buckwheat noodles was developed by using the fluorescence fingerprint and partial least squares regression. Fitting the calibration model to validation datasets showed R(2)=0.78 and SEP=12.4%. The model was refined for a better fit by deleting several samples containing additional ingredients. The best fit was finally obtained (R(2)=0.84 and SEP=10.4%) by deleting the samples containing vinegar, green tea, seaweed, polysaccharide thickener, and yam. This result demonstrates that a calibration model with high accuracy could be constructed based on samples similar in material composition. The developed methodology requires no complex preprocessing, enables rapid measurement with a small sample amount, and would thus be suitable for practical application to the food industry.