Established Prediction Models of Bread Sensory Evaluation Results Based on Data Analysis Methods

Abstract

In this paper, 66 samples belonging to 12 varieties of domestic high-gluten wheat in 2021 were collected, and their physicochemical, rheological and bread sensory properties were analyzed. The results of the main component analysis showed that the total sensory score of bread had a strong correlation with gluten index, maximum tensile resistance, mixing time, stretching area and stabilization time, etc. At the same time, the screening of wheat varieties with high similarity through cluster analysis could play a key guiding role in the development and maintenance of production stability of baking special flour. Comparing the fitting quality of the bread total sensory score prediction model established by the three methods of stepwise regression, partial least squares and neural network model, the model built using artificial neural network was significantly better than other models. Using neural network models, the baking characteristics of different varieties of wheat could be quickly predicted, ensuring product stability, while also facilitating the development of more targeted baking special flours.

Keywords

• sensory evaluation
• principal component analysis
• cluster analysis
• neural network
• predictive modeling