随机多粒度标记序决策系统的最优粒度选择  

Optimal Granularity Selections for Random Multi-GranularLabeled Ordered Decision Systems

作　　者：林玉梅 方连花 吴伟志 曹飞龙 LIN Yumei;FANG Lianhua;WU Weizhi;CAO Feilong(School of Software,Quanzhou University of Information Engineering,Quanzhou 362000,Fujian,China;GeneralEducation Center,Quanzhou University of Information Engineering,Quanzhou 362000 Fujan,China;School ofInformation Engineering,Zhejiang Ocean University,Zhoushan 316022,Zhejiang,China;School of Electronicsand Communication Engineering,Quanzhou University of Infomation Engineering,Quanzhou 362000,FujianChina;College of Sciences,China Jiliang University,Hangzhou 310018,China)

机构地区：[1]泉州信息工程学院软件学院,福建泉州362000 [2]泉州信息工程学院通识中心,福建泉州362000 [3]浙江海洋大学信息工程学院,浙江舟山316022 [4]泉州信息工程学院电子与通信学院,福建泉州362000 [5]中国计量大学理学院,浙江杭州310018

出　　处：《昆明理工大学学报（自然科学版）》2023年第4期194-204,共11页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(61976194,62076221);福建省自然科学基金项目(2021J01540);福建省中青年教师教育科研项目(JAT210556);福建省职业教育教学改革研究课题(GB2020036)。

摘　　要：传统随机信息系统的属性一般只有一个粒度层面,即系统中的每一个对象在每一个属性上取唯一的值,而在实际应用中,人们往往需要从不同的粒度层面去获取知识.实际生活中数据集的一个对象在同一个属性上根据不同的粒度可以取不同值.针对多粒度数据集的知识获取问题,本文讨论随机多粒度标记序信息(决策)系统的最优粒度选择问题.首先,通过优势关系给出随机多粒度标记序信息系统的信息粒表示,进一步定义由条件属性集导出集合的下近似与上近似,描述在不同标记下的信息粒的结构以及上、下近似的变化关系;其次,用证据理论的信任函数与似然函数来刻画随机多粒度标记序决策系统的两种最优粒度;最后,给出协调和不协调随机多粒度标记序决策系统的最优粒度选择算法.Atributes in a traditional random information system have only one level of granularity,meaning thateach object under each atribute in the system can only have one value.However,in practical applications,it isoften necessary to acquire knowledge from different granularity levels.In many real-life datasets,an object canhave multiple values under the same attribute,corresponding to different scales.To acquire knowledge in datasetswith multiple granularity levels,this study discusses the optimal scale selection in random multi-granularity la-beled ordered information/decision systems.Firstly,information granules in random multi-granularity labeled or-dered information systems are represented using dominance relations.Lower and upper approximations of sets,generated by condition atribute subsets with respect to dominance relations,are defined.The structures of infor-mation granules,as well as the upper and lower approximations with dfferent granular labels,are also described.Furthermore,the Dempster-Shafer theory of evidence is employed to characterize two types of optimal scales inrandom multi-granularity labeled ordered decision systems:belief and plausibility functions.Finally,two algo-rithms for optimal scale selection are designed for consistent and inconsistent random multi-granularity labeledordered decision systems,respectively.

关 键 词：粒计算 多粒度标记 随机序决策系统 信任函数 最优粒度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]