首页> 外国专利> How to create a two-class classification prediction model, creating apparatus of two-class classification and predictive models of the program for the classification prediction model created

How to create a two-class classification prediction model, creating apparatus of two-class classification and predictive models of the program for the classification prediction model created

机译:如何创建两类分类预测模型,如何创建两类分类预测模型的程序以及用于该类预测模型的程序

摘要

Regardless of the difference in classification method, by performing a two-class classification nearly 100% substantially, to create a two-class classification / prediction model reliable. The method of the present invention, the steps of preparing first, by performing a discriminant analysis to the sample set as the training data sample set comprising a plurality of samples belonging to the class of the second sample belonging to the first class and a second step of creating a determined second function with a discriminant function first with high classification characteristics for the first class, high classification characteristics for the second class, discriminant function of the first and second Run the classification of the sample set was used to using a new sample set and the third step of identifying, samples identified samples classification results of the two do not match, repeating steps a second, third If the fourth step, the number of samples that do not match the in the third step becomes a predetermined value or less, and a step of a fifth stopping the fourth step.
机译:不管分类方法上的差异如何,通过实质上执行近100%的两类分类,来创建可靠的两类分类/预测模型。本发明的方法,首先准备步骤,通过对作为训练数据样本集的样本集进行判别分析,该训练数据样本集包括属于第一类的第二样本的多个样本和第二步骤首先创建具有判别函数的确定的第二函数,该函数具有第一类的高分类特征,第二类的高分类特征,第一类和第二类的判别函数运行样本集的分类用于使用新的样本集并且,在第三识别步骤中,将被识别出的两个样本的分类结果不匹配的样本,重复第二,第三步骤,如果第四步骤中,第三步骤中不匹配的样本数为规定值以下,第五步停止第四步。

著录项

  • 公开/公告号JPWO2008059624A1

    专利类型

  • 公开/公告日2010-02-25

    原文格式PDF

  • 申请/专利权人 富士通株式会社;

    申请/专利号JP20080544075

  • 发明设计人 湯田 浩太郎;

    申请日2007-03-27

  • 分类号G06N5/04;G06N3/00;G06F17/30;

  • 国家 JP

  • 入库时间 2022-08-21 18:57:01

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号