多粒度层次序贯三支决策模型研究  

Research on multi⁃granularity hierarchical sequential three⁃way decision model

作　　者：钱进 汤大伟 洪承鑫 QIAN Jin;TANG Da-wei;HONG Cheng-xin(School of Computer,Jiangsu University of Science and Technology,Zhenjiang 212100,Jiangsu,China;School of Software,East China Jiaotong University,Nanchang 330013,Jiangxi,China)

机构地区：[1]江苏科技大学计算机学院,江苏镇江212100 [2]华东交通大学软件学院,江西南昌330013

出　　处：《山东大学学报（理学版）》2022年第9期33-45,共13页Journal of Shandong University(Natural Science)

基　　金：国家自然科学基金资助项目(62066014);江西省自然科学基金资助项目(20202BABL202018);江苏省“青蓝工程”中青年学术带头人培养项目;江西省“双千计划”项目。

摘　　要：现有的知识获取算法所挖掘出的规则太多,不易理解;规则描述太过具体,容易造成过拟合。为此,本文提出了多粒度层次序贯三支决策模型。首先引入概念层次树将目标概念泛化,构建多层次决策表,并设计了多粒度层次序贯三支决策模型,从多视角、多层次计算3个概率区域并获取相应的泛化层次决策规则。最后,通过实验证明了模型的有效性。本模型为知识获取提供了新的视角并丰富了多粒度三支决策的研究。The existing algorithms generate a large number of the decision rules,which is not easy to be understood.Moreover,the descriptions of the rules are too specific,which easily lead to overfitting.To this end,a multi⁃granularity hierarchical sequential three⁃way decision model is proposed.Firstly,we generalize the target concept through the concept hierarchy tree and construct the multi⁃hierarchical decision table.Then,we construct the multi⁃granularity hierarchical sequential three⁃way decision model,calculate the three probabilistic regions,and acquire the corresponding the generalized rules from the multi⁃view and multi⁃level.Finally,the effectiveness of our model is verified by experiments.This model provides a new perspective for knowledge acquisition and enriches the research of multi⁃granularity three⁃way decisions.

关 键 词：知识获取 三支决策 多粒度粗糙集 多层次决策表 概念层次树 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]