PROGRAM FOR CONSTRUCTING EXPANSION WEIGHT UPDATE TYPE SELF-ORGANIZATION MAP, PROGRAM FOR ESTIMATING CHARACTERISTIC VALUE USING CONSTRUCTED EXPANSION WEIGHT UPDATE TYPE SELF-ORGANIZATION MAP AND CHARACTERISTIC VALUE ESTIMATION APPARATUS EMPLOYING EXPANSION WEIGHT UPDATE TYPE SELF-ORGANIZATION MAP

摘要

PROBLEM TO BE SOLVED: To provide a program for constructing an expansion weight update type self-organization map while using much data for learning.SOLUTION: When a feature amount obtained by measuring a substance is inputted, a program for constructing an expansion weight update type self-organization map for quantitatively calculating and outputting a characteristic value of which the character is different from the feature amount, allows a computer to execute the steps of accepting input of a plurality of measurements of the feature amount and a plurality of measurements of the characteristic value which are measured under different conditions (step S1), and calculating variance of the measurements and excludes the measurements of which the variance is equal to or more than a predetermined threshold value (step S3). Further, outlier values are excluded from among the measurements (step S4) and a learning sample is created by making the remaining measurements of the feature amount and the characteristic value through statistical analysis (step S5). In accordance with the created learning sample, the self-organization map is allowed to perform expansion weight update learning so as to be constructed as an expansion weight update type self-organization map having a competitive layer in which a relationship between feature amounts is mapped (step S6).