MODELING SENSORY COLOR QUALITY OF TOMATO AND PEACH: NEURAL NETWORKS AND STATISTICAL REGRESSION

Abstract

ABSTRACT In this study, neuro-computing and statistical regression techniques were applied to the determination of mathematical relationships linking human sensory judgments to physical measurements of external color for tomato and peach. Both techniques provided the same answers regarding the selection of the factors to be included in the final mathematical equations. Statistical techniques did have a slight advantage in numerical accuracy as shown by paired-comparison t-tests, but neuro-computing techniques involved less steps during the analysis phase. Thus, a compromise recommendation is to use neural networks to get the basic features of the linkages, then to use statistical techniques to fine tune the numerical accuracy of resulting mathematical functions. Relationships between sensory evaluations and physical measurements of tomato and peach external color were found to be mostly linear.