文献类型：期刊文献

中文题名：基于层次结构的多分类算法研究

英文题名：Research on Multiclass Classification Based on Hierarchical Model

作者：任进军[1];陈军[1]

第一作者：任进军

机构：[1]甘肃中医药大学理科教学部,甘肃定西743000

第一机构：甘肃中医药大学定西校区

年份：2017

卷号：34

期号：4

起止页码：59

中文期刊名：贵州大学学报（自然科学版）

外文期刊名：Journal of Guizhou University:Natural Sciences

收录：CSTPCD

基金：甘肃中医药大学定西校区校级项目(2017XJYB05;2017XJYB08;TD2016ZD05)

语种：中文

中文关键词：多分类;算法;层次模型;研究

外文关键词：muhiclass classification; algorithm; hierarchical model ; study

摘要：在众多多分类问题的解决方法中,分解策略因其简单易行而成为主要的解决方法,本文对决策有向无环图(Decision Directed Acyclic Graphs,DDAG)和有向二叉树(Directed Binary Tree,DBT)两种层次模型从结构特点、算法执行时间以及分类准确率三个方面进行分析,为模型的择优使用提供参考。
The decomposition strategy is a main and simple solution to solve the problem,and the Decision Directed Acyclic Graphs model and Directed Binary Tree model are the most popular tactics. It is difficult to accurately grasp and distinguish,because the structure and organization of two hierarchical models are similar. The multiclass classification of the models was discussed in views of structural characteristics,execution time and classification accuracy,and some suggestion was given.