摘要
粒计算是数据挖掘和知识发现的一个重要手段,以信息粒作为基本计算单位,可以有效地处理大规模数据.不完备序决策系统拓展了经典粗糙集模型,将优势关系引入粗糙集模型使其能够更好地处理带有偏好关系的不完备决策表.本文结合了粗糙集和证据理论,研究了在不完备序决策系统中,序上、下近似算子和证据理论中的信任函数和似然函数的关系,给出了如何根据序上、下近似算子来计算信任函数和似然函数的方法.进而提出了在不完备序决策系统中证据理论下,近似分布约简和相对信任约简的定义和相关性质,并研究了它们之间的一致性,为粒计算提供了一种新的思路.
An important task to deal with large amount of data is to establish relations among granules. Incomplete ordered decision system expands the classic rough sets theory by applying the dominance relation instead of the equivalence relation,which makes it capable of dealing with incomplete data with preference relationship. This paper explains the relationship between the approximations with the belief functions,and provides a method to acquire belief functions based on the approximations in the incomplete decision system.The definition and characteristics of approximation reduction and belief reduction are given and the consistency features are studied.Our work provides a new method to describe the attribute reduction in the incomplete ordered decision system.
作者
陈一舟
王加阳
郑娜
CHEN Yi-zhou;WANG Jia-yang;ZHENG Na(Department of Information Science and Engineering,Central South University,Changsha 410083,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第12期2566-2570,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61772031)资助
长沙市2017年节能专项资金项目资助
湖南省自然科学基金项目(2018JJ2131)资助
中南大学中央高校基本科研业务费专项基金项目(2017zzts514)资助
关键词
粒计算
不完备序决策系统
证据理论
属性约简
granular computing
incomplete ordered decision system
evidence theory
attribute reduction
