Characteristic Collection of Taste Sensor Based on Sensory Evaluation and its Application to Food Discrimination

Abstract

A function that expresses the relationship between data obtained from a sensory evaluation and those obtained from a taste sensor was approximated using a polynomial. First, the basis functions of a polynomial were set to a Legendre function, and the number of basis functions and their degrees were selected using a genetic algorithm. Next, the coefficients of the polynomial were calculated using multiple regression analysis. The approximation error was estimated using cross-validation, and the number of basis functions and their degrees were optimized so as to maximize the generalization of the polynomial. As a result, sufficiently small approximation errors were obtained for test data, and the explicit relationship between the data obtained from a sensory evaluation and those obtained from a taste sensor were clarified using the polynomial. The discrimination rate for food was improved using the estimated data of a sensory evaluation compared to a method that used only the data obtained from a taste sensor.