TWO-CLASS CLASSIFYING/PREDICTING MODEL MAKING METHOD, CLASSIFYING/PREDICTING MODEL MAKING PROGRAM, AND TWO-CLASS CLASSIFYING/PREDICTING MODEL MAKING DEVICE

摘要

A highly-reliable two-class classifying/predicting model is made by performing substantially 100% two-class classification independently of the difference of classification scheme. The method of the invention comprises a first step of preparing as learning data a sample set containing samples belonging to a first class and samples belonging to a second class, a second step of performing judgment analysis of the sample set to define a first judgment function having a high classification characteristic for the first class and a second judgment function having a high classification characteristic for the second judgment function, a third step of classifying the samples of the sample set by using the first and second judgment functions and detecting samples for which the results of both classifications disagree with each other, a fourth step of repeating the second and third steps by using the detected samples as a new sample set, and a fifth step of stopping the fourth step if the number of samples for which the results disagree at the third step becomes a given value or less.