首页> 外国专利> MULTIPLE-CLASSIFIER INTEGRATED CONTROL SYSTEM AND METHOD THEREOF

MULTIPLE-CLASSIFIER INTEGRATED CONTROL SYSTEM AND METHOD THEREOF

机译:多分类器综合控制系统及其方法

摘要

In a multi-classifier integrated control method performed by a multi-classifier integrated control system for classifying a classification object according to an aspect of the present application, basic information on classification algorithm and structure information applied to each of the multi-classifiers is collected and collected. Loading a plurality of classifiers according to the classification target based on the basic information; When a classification target is input to a plurality of loaded classifiers, providing at least one classification item through a classification process for each classifier; Among the classification items provided in each classifier, a classification is made by classifying duplicate items having duplicate classification items and unique items having non-overlapping classification items, and setting weight functions for converging the weight and prediction result reliability for duplicate items. Preparatory stage; It collects the input information for the classification target, pre-processes the input data to be input to the classifier, assigns the pre-processed information to a plurality of classifiers, performs the classification process, and calculates the weight of the classification items and the reliability of prediction results Classification performance step; And an integrated convergence step of converging the weight of the duplicate items calculated by the classifier using the set weight function and the reliability of the prediction result into one value. The loaded classifier is a classifier that outputs weights and prediction result reliability for classification items.
机译:在根据本申请的一方面的由用于将分类对象分类的多分类器集成控制系统执行的多分类器集成控制方法中,收集关于分类算法的基本信息和应用于每个多分类器的结构信息,并且集。根据基本信息根据分类目标加载多个分类器;当分类目标输入到多个加载的分类器时,通过分类过程为每个分类器提供至少一个分类项目;在每个分类器中提供的分类项目中,通过对具有重复的分类项目的重复项目和具有不重叠的分类项目的唯一项目进行分类,并设置权重函数以对重复项目的权重和预测结果可靠性进行收敛,来进行分类。准备阶段;它收集分类目标的输入信息,对要输入到分类器的输入数据进行预处理,将预处理后的信息分配给多个分类器,执行分类处理,并计算分类项目和分类的权重。预测结果的可靠性分类性能步骤;以及集成的收敛步骤,其将使用设置的权重函数将分类器计算出的重复项的权重和预测结果的可靠性收敛为一个值。加载的分类器是输出分类项目的权重和预测结果可靠性的分类器。

著录项

  • 公开/公告号KR102133972B1

    专利类型

  • 公开/公告日2020-07-15

    原文格式PDF

  • 申请/专利权人 세종대학교산학협력단;

    申请/专利号KR20190089020

  • 发明设计人 한동일;우현준;김미선;

    申请日2019-07-23

  • 分类号G06K9/62;G06N20;

  • 国家 KR

  • 入库时间 2022-08-21 11:04:16

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