Prediction of consumer acceptance in some thermoprocessed food by physical measurements and multivariate modeling

Abstract

Consumer acceptances for French bread, fish bread, and roasted coffees were calibrated against physical measurements of those products using Multiple Linear Regression. The models obtained were then validated and tested using the widespread used methods of cross-validation, y-randomization and external validation. In all cases, multivariate models presented R2 for calibration greater than.9, which was superior to those univariate ones. For the French bread analysis, the multivariate model performed well and the length of the cut on bread surface is the parameter that most strongly influenced this model; on the other hand, a large width of the cut on bread surface would greatly contribute to a lower acceptance. The model for predicting the acceptance of the fish bread also showed a good performance; the bulkier fish breads received a better acceptance. An efficient model was also obtained for the data set of roasted coffee; redder coffees were more accepted. Practical applications A multivariate regression was used in order to predict the consumer acceptance from measurements commonly performed for characterization of food products. Consumer acceptance can be predicted by easy and rapid physical (and/or chemical) measurements using regression models. Once built and validated, the models can be used to predict the consumer acceptance by rapid physical measurements on the products. This approach can be a useful method to be included as a quality control parameter on food industry.