南京大学学报(自然科学版), 2019, 55(4): 553-563 doi: 10.13232/j.cnki.jnju.2019.04.005

基于条件属性蕴含的概念格构造及简化

王霞,1,2, 谭斯文1, 李俊余1,2, 吴伟志1,2

1. 浙江海洋大学数理与信息学院,舟山,316022

2. 浙江省海洋大数据挖掘与应用重点实验室,浙江海洋大学,舟山,316022

Constructions and simplifications of concept lattices based onconditional attribute implications

Wang Xia,1,2, Tan Siwen1, Li Junyu1,2, Wu Weizhi1,2

1. School of Mathematics,Physics and Information Science,Zhejiang Ocean University,Zhoushan,316022,China

2. Key Laboratory of Oceanographic Big Data Mining and Application of Zhejiang Province,Zhejiang Ocean University,Zhoushan,316022,China

通讯作者: E⁃mail:bblylm@126.com

收稿日期: 2019-05-28   网络出版日期: 2019-07-17

基金资助: 浙江省自然科学基金.  LY18F030017

Received: 2019-05-28   Online: 2019-07-17

摘要

基于三元背景研究三类概念格的构造和简化.首先,基于三元背景构造一个条件属性蕴含形式背景,该背景以三元背景属性集上的属性蕴含为对象,以三元背景的条件为属性.并针对条件属性蕴含形式背景给出形式概念的定义,构造相应的概念格.其次,由于条件属性蕴含形式背景中对象的个数随着三元背景中属性个数的增加呈指数级增长,这使得条件属性蕴含形式背景往往是一个比较大的数据表,因此,对条件属性蕴含形式背景进行对象约简,将原来的对象集替换为单个条件下形式背景的极小属性蕴含构成的集合.该对象约简方法不仅在很大程度上简化了条件属性蕴含形式背景,而且简化后的形式背景对应的概念格与原来的概念格同构.最后,在条件属性蕴含形式背景上引入了可能性算子和必然性算子,在此基础上定义了对象定向概念格和属性定向概念格.

关键词: 概念格 ; 三元背景 ; 条件属性蕴含 ; 对象定向概念格 ; 属性定向概念格

Abstract

Constructions and simplifications of three types of concept lattice are studied based on a triadic context. Firstly,a new formal context is constructed based on conditional attribute implications,which takes the implications between attributes of the triadic context as the objects and the conditions of the triadic context as the attributes. Then definitions of formal concept and concept lattice are given in the conditional attribute implication context. Secondly,since the number of objects in the conditional attribute implication context increases exponentially with the increase of the number of attributes in the triadic context,which makes the conditional attribute implication context usually becomes a large data table. The object reduction of the conditional attribute implication context is carried out,and the original object set is replaced with the set of minimal attribute implications of the formal context under each single condition. It is shown that the object reduction method can simplify the conditional attribute implication context to a great extent,and the concept lattice corresponding to the simplified context is isomorphic to the original concept lattice. Finally,the possibility operator and necessary operator are introduced in the conditional attribute implication context to define the object oriented concept lattice and property oriented concept lattice of the new context.

Keywords: concept lattice ; triadic context ; conditional attribute implication ; object oriented concept lattice ; property oriented concept lattice

PDF (543KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

王霞, 谭斯文, 李俊余, 吴伟志. 基于条件属性蕴含的概念格构造及简化. 南京大学学报(自然科学版)[J], 2019, 55(4): 553-563 doi:10.13232/j.cnki.jnju.2019.04.005

Wang Xia, Tan Siwen, Li Junyu, Wu Weizhi. Constructions and simplifications of concept lattices based onconditional attribute implications. Journal of nanjing University(Natural Science)[J], 2019, 55(4): 553-563 doi:10.13232/j.cnki.jnju.2019.04.005

三元概念分析是由Lehmann and Wille[1,2]于1995年提出的,它可以看作是形式概念分析[3,4]的一种自然扩展.目前,三元概念分析主要集中在三元概念和概念三元格的构造、三元蕴含及关联规则挖掘、三元模态算子、三元概念聚类、三元背景的因子分析、模糊化、三元概念模型推广和约简及应用等方面,取得了一些研究成果.Wei et al[5,6]综述了有关三元概念分析的基本理论、方法及应用等相关研究内容.

属性蕴含是描述某些属性依赖关系的特定表达式,是形式概念分析的重要研究课题之一.1982年Ganter and Wille[4]定义形式背景上属性蕴含AB表示每个具有A中属性的对象一定具有B中属性.近年,属性蕴含相关的问题引起了许多学者的广泛关注,取得了一些研究成果[7,8,9,10,11,12,13,14,15].1998年Biedermann[16]将属性蕴含的思想引入三元概念分析,定义了在某个条件集下的属性间的三元蕴含.2004年Ganter and Obiedkov[17]定义了一种新的条件属性蕴含并进一步定义了属性条件蕴含和属性×条件蕴含,基于条件属性蕴含构造一个形式背景,给出了形式概念的描述性定义.Glodeanu[18]定义了一种模糊值三元背景,并定义了模糊值条件属性蕴含、模糊值属性条件蕴含和模糊值属性×条件蕴含.Mora et al[19]和Rodríguez⁃Lorenzo[20]研究了三元概念分析中条件属性蕴含的公理系统和推理系统.

本文受Ganter and Obiedkov[17]的启发,基于一个三元背景构造条件属性蕴含形式背景,并定义该背景的形式概念,构造相应的概念格.考虑到条件属性蕴含形式背景中对象的个数随着三元背景中属性个数的增长呈指数级增长,因此在实际应用中条件属性蕴含形式背景往往是一个比较大的数据表.本文对条件属性蕴含形式背景的对象进行了约简,从而使形式背景和形式概念的表示更加简洁,并且该约简方法可使简化后的形式背景对应的概念格与原来的概念格同构.最后,在条件属性蕴含形式背景中引入可能性算子和必然性算子,构造了对象定向概念格和属性定向概念格.

1 相关理论

1.1 形式概念分析相关知识

定义1[3]T:=(G,M,I)为一个形式背景,其中,G是一个对象集,M是一个属性集,IGM之间的一个关系.分别称GM的元素为对象和属性.

若对象g和属性m具有关系I,则记为(g,m)IgIm.

定义2[3]T=(G,M,I)为形式背景,XG,BM.若二元组(X,B)满足X'=B,B'=X,则称(X,B)为形式概念,其中:

X'={mM|gX,(g,m)I}
B'={gG|mB,(g,m)I}

T=(G,M,I)是形式背景,对任意的形式概念(X1,B1),(X2,B2)定义如下偏序关系:

(X1,B1)(X2,B2)X1X2(B1B2)

L(G,M,I)L(T)T=(G,M,I)中所有形式概念构成的集合,则L(T)是格,称其为T的概念格.在概念格L(T)上定义上、下确界如下:

(X1,B1)(X2,B2)=X1X2,(B1B2)
(X1,B1)(X2,B2)=(X1X2),B1B2

则概念格L(T),,是一个完备格.

gG,mM,分别称({g},{g}')({m}',{m})为对象概念和属性概念.

定义3[3]T=(G,M,I)是形式背景,A,BM.称二元有序对(A,B)为集合M上的属性蕴含,记为AB,读作“A蕴含B”.分别称AB为属性蕴含AB的前件和结论.称属性蕴含ABT中成立,若它满足条件:gG,g具有A中所有属性,则g具有B中所有属性.

BAM,则属性蕴含ABT中成立,此时称属性蕴含AB是平凡的.若A'=,则称属性蕴含AB的前件是假的.

根据定义3,A,BM,属性蕴含ABT中成立当且仅当mB,A{m}T中成立.

因此,下文只考虑A{m}形式的非平凡属性蕴含,简记为Am.Imp(T)为在形式背景T中成立的非平凡属性蕴含构成的集合,即Imp(T)={Am|mM\A,AmT

} .

1.2 三元概念分析相关知识

定义4[1]K=(K1,K2,K3,Y)为三元背景,其中K1,K2,K3为非空集合,YK1,K2,K3之间的关系,即YK1×K2×K3.分别称K1,K2,K3为对象集、属性集和条件集.分别称K1,K2,K3的元素为对象、属性和条件.

若对象g、属性m和条件c具有关系Y,则记为(g,m,c)Y,表示对象g在条件c下具有属性m.

K=(K1,K2,K3,Y)是一个三元背景,A1K1,A2K2,A3K3,称三元组(A1,A2,A3)为一个三元概念,若满足A1×A2×A3YX1K1,X2K2,X3

K3,X1×X2×X3Y,当A1X1,A2X2,A3X3时总有(A1,A2,A3)=(X1,X2,X3).此时分别称A1A2A3为三元概念(A1,A2,A3)的外延、内涵和模式.

由定义1、定义2和定义4可知,三元背景和三元概念分别是形式背景和形式概念的扩展,因此形式概念分析和三元概念分析在研究方法和研究内容等方面都有着紧密的联系[21,22,23,24,25].形式概念分析的研究对象是形式背景,它是一个二维数据表,当遇到三维数据或基于条件下的二维数据时,三元概念分析则有效地扩展和提升了形式概念分析的信息处理能力.目前,三元概念分析在folksonomy分类[26,27,28]、认知系统[29]、文本分类[30]、联盟应用[31]、动态社会网络[32]、访问控制[33]等方面也取得了一些研究成果.

定义5[17]K=(K1,K2,K3,Y)为三元背景,A,BK2,CK3.ACB为基于三元背景的条件属性蕴含,读作“在C中所有条件下A蕴含B”.称条件属性蕴含ACB在三元背景K中成立,若它满足条件:在每一个条件cC下,当gK1,g具有A中所有属性时,g具有B中所有属性.

定义6K=(K1,K2,K3,Y)为三元背景,CK3.KC:=(K1,K2,YC)为在条件子集C下的形式背景,其中gK1,mK2,(g,m)YC当且仅当cC,(g,m,c)Y.

BAK2,则条件属性蕴含ACBK中成立,此时称ACB是平凡的.若A'C=,则称ACB的前件是假的.

根据定义5可得,条件属性蕴含ACBK中成立当且仅当mB,AC{m}K中成立.

类似的,以下只考虑AC{m}形式的条件属性蕴含,将其简记为ACm,并记:

ImpC(K):={ACm|AK2,mK2\A,CK3,ACmK}

例1表1给出了一个三元背景K=(K1,K2,K3,Y),其中对象集K1={g1,g2},属性集K2={m1,m2,m3},条件集K3={c1,c2,

表1   三元背景

Table 1  A triadic context

c1c2c3
m1m2m3m1m2m3m1m2m3
g1101011101
g2010110110

新窗口打开| 下载CSV


c3}.根据定义6可得在条件c1下的形式背景Kc1=(K1,K2,Yc1),如表2所示.

表2   在条件c1下的形式背景Kc1

Table 2  The formal context Kc1 under the condition c1

m1m2m3
g1101
g2010

新窗口打开| 下载CSV


根据定义2可知,表2中形式背景Kc1有四个形式概念,分别为:(K1,),({g1},{m1,m3}),({g2},{m2}),(,K2)

根据定义3可得,在形式背景Kc1中成立的非平凡属性蕴含集为:

Imp(Kc1)=m1m3,m3m1,{m1,m2}m3,{m2,m3}m1

取条件子集C={c1,c3},则根据定义5可得在三元背景K中成立的非平凡条件属性蕴含集为:

ImpC(K2)={m3Cm1,{m2,m3}Cm1}

2 基于条件属性蕴含的概念格

定义7K=(K1,K2,K3,Y)为三元背景,记imp(K2)={Am|AK2,mK2\A},称形式背景Cimp(K)=(imp(K2),K3,R)为基于K的条件属性蕴含形式背景,其中Amimp(K2),cK3,(Am,c)R当且仅当AcmImpc(K).

条件属性蕴含形式背景Cimp(K)的对象为三元背景K的属性集K2上的属性蕴含,属性为三元背景K的条件集K3中的条件.

定义8K=(K1,K2,K3,Y)为三元背景,Cimp(K)=(imp(K2),K3,R)为基于K的条件属性蕴含形式背景,imp(K2),CK3.若二元组(,C)满足:'=C,C'=,则称(,C)为条件属性蕴含形式背景Cimp(K)的形式概念,此时分别称C为形式概念(,C)的外延和内涵.其中,

'={cK3|Am,(Am,c)R}
C'={Amimp(K2)|cC,(Am,c)R}

对任意的概念(1,C1),(2,C2)定义偏序关系:

(1,C1)(2,C2)12(C1C2)

L(Cimp(K))Cimp(K)中所有概念构成的集合,在L(Cimp(K))上定义上、下确界:

(1,C1)(2,C2)=(12,(C1C2))
(1,C1)(2,C2)=((12),C1C2)

(L(Cimp(K)),,)是一个完备格,称其为条件属性蕴含形式背景Cimp(K)的概念格.

根据定义3、定义5和定义6知,Am

Imp(KC)当且仅当ACmImpC(K).

K=(K1,K2,K3,Y)为三元背景,Cimp(K)为条件属性蕴含形式背景,由定义8知,cK3,{c}'=Imp(Kc).

(,C)L(Cimp(K)).C,则根据式(4)知:

={{c}'|cC}={Imp(Kc)|cC}

例2表1对应的条件属性蕴含形式背景Cimp(K)=(imp(K2),K3,R)表3所示,其中:

表3   条件属性蕴含形式背景Cimp(K)

Table 3  The conditional attribute formal context Cimp(K)

c1c2c3
m1001
m2010
m3000
m1m2010
m1m3100
m2m1001
m2m3000
m3m1101
m3m2010
{m1,m2}m3100
{m1,m3}m2010
{m2,m3}m1101

新窗口打开| 下载CSV


imp(K2)=m1,m2,m3,m1m2,m1m3,m2m1,m2m3,m3m1,m3m2,{m1,m2}m3,{m1,m3}m2,{m2,m3}m1

首先,计算条件属性蕴含形式背景Cimp(K)的三个属性概念如下:

C1:=({c1}',{c1}'')=({m1m3,m3m1,{m1,m2}m3,{m2,m3}m1},{c1}
C2:=({c2}',{c2}'')=({m2,m1m2,m3m2,{m1,m3}m2},{c2}
C3:=({c3}',{c3}'')=({m1,m2m1,m3m1,{m2,m3}m1},{c3})

其次,根据式(4)和式(5)计算剩余的三个概念:

C4:=({c1}',{c1}'')({c3}',{c3}'')=({m3m1,{m2,m3}m1},{c1,c3})
C5:=({c1}',{c1}'')({c2}',{c2}'')({c3}',{c3}'')=(,K3)
C6:=({c1}',{c1}'')({c2}',{c2}'')({c3}',{c3}'')=(imp(K2),)

所有概念对应的概念格L(Cimp(K))图1所示.

图1

图1   概念格L(Cimp(K))

Fig.1   The concept lattice L(Cimp(K))


若三元背景K=(K1,K2,K3,Y)的属性集K2n(n1)个属性,则条件属性蕴含背景Cimp(K)的对象集imp(K2)n2n-1个元素,因此随着n的增大imp(K2)的元素个数呈指数级增长.例如当n=5时,imp(K2)有80个元素;当n=6时,imp(K2)有192个元素……鉴于此,下面考虑对条件属性蕴含背景Cimp(K)的对象做适当的约简.

T=(G,M,I)是形式背景,根据定义3知,AmImp(T),BM,若AB,则BmImp(T).

定义9T=(G,M,I)是形式背景,AmImp(T)总存在A0A满足:A0mImp(T)A1A,若A1A0,则A1mImp(T),此时称A0m为关于Am的极小属性蕴含.

MinImp(T)为形式背景T的所有极小属性蕴含构成的集合.显然有:

Imp(T)=Am|A0mMinImp(T)使A0A

定义10K=(K1,K2,K3,Y)为三元背景,Cimp(K)=(imp(K2),K3,R)为基于K的条件属性蕴含形式背景.

SCimp(K)=(Simp(K2),K3,RSimp(K2))Cimp(K)的简化背景,其中,

Simp(K2)={Am|cK3,AmMinImp(Kc)}

AmSimp(K2),cK3,(A

m,c)RSimp(K2)当且仅当AcmImpc(K).

显然有Am,cRSimp(K2)当且仅当AmImp(Kc)Simp(K2).

根据定义10知:

Simp(K2)={MinImp(Kc)|cK3}
Simp(K2) ,CK3

记:

'Simp(K2):={cK3|Am,(Am,c)RSimp(K2)}='
C'Simp(K2):={AmSimp(K2)|cC,(Am,c)RSimp(K2)}=C'Simp(K2)

对比定义7和定义10可知,条件属性蕴含形式背景Cimp(K)和简化背景SCimp(K)的对象集不同,而且简化背景SCimp(K)的对象集Simp(K2)所包含的元素个数明显比原背景中对象集imp(K2)少得多.事实上,设cK3使AmMinImp(Kc),且不妨设|A|=s,则根据定义9可知,BK2,若BAABBmMinImp(Kc)Bmimp(Kc).而这样的属性子集共有(2s+2n-s-1-2)个,而且可以验证当s=n-12时,2s+2n-s-1-2=22n-12-1达到最小值.即若一个极小属性蕴含的前件含有s个属性,则可以从imp(Kc)中剔除掉(2s+2n-s-1-2)个属性蕴含,且至少剔除掉22n-12-1个,因此简化后的背景SCimp(K)在很大程度上降低了对象的个数.

引理1K=(K1,K2,K3,Y)为三元背景,Cimp(K)为条件属性蕴含背景,SCimp(K)Cimp(K)的简化背景,imp(K2),CK3.(,C)L(SCimp(K)),则:

(Simp(K2),C)L(SCimp(K))

明 因为(,C)L(SCimp(K)),则:

C'Simp(K2)=C'Simp(K2)=Simp(K2)

又因为:

(Simp(K2))'Simp(K2)=(Simp(K2))'

所以有:

'(Simp(K2))'=(Simp(K2))'Simp(K2)

另一方面,若Simp(K2)=,则:

(Simp(K2))'Simp(K2)=(Simp(K2))'='=C

于是:

(Simp(K2),C)L(SCimp(K))

Simp(K2),则:Am\Simp(K2)存在A0K2满足A0AA0mSimp(K2).

因此:

c(Simp(K2))'Simp(K2)=(Simp(K2))'

A0cmImpc(K).

所以:

AcmImpc(K) ,即c(\Simp(K2))'

于是,

(Simp(K2))'(\Simp(K2))'

即:

(Simp(K2))'='=C

所以:

(Simp(K2),C)L(SCimp(K))

证毕.

定理1K=(K1,K2,K3,Y)为三元背景,Cimp(K)为条件属性蕴含背景,SCimp(K)Cimp(K)简化背景,则概念格L(Cimp(K))与概念格L(SCimp(K))同构.

明 设:

f(,C)=(Simp(K2),C),(,C)L(Cimp(K)),则由引理1知fL(Cimp(K))L(SCimp(K))的映射,且为单射.

(,C)L(SCimp(K)) ,令:
1=Bm|Am使ABmB

则有:

(1,C)L(Cimp(K))

f((1,C))=(1Simp(K2),C)=(,C)

fL(Cimp(K))L(SCimp(K))的满射.

于是,fL(Cimp(K))L(SCimp(K))的双射.

此外,(1,C1),(2,C2)L(Cimp(K))

f((1,C1)(2,C2))=f(12,(C1C2))=(12Simp(K2),(C1C2))=(1Simp(K2),C1)(2Simp(K2),C2)=f((1,C1))f((2,C2))
f((1,C1)(2,C2))=f((12),C1C2)=((12)Simp(K2),C1C2)=(1Simp(K2),C1)(2Simp(K2),C2)=f(1,C1)f(2,C2)

则,概念格L(Cimp(K))L(SCimp(K))同构.

证毕.

例3(续例2) 表4给出了例2中条件属性蕴含背景Cimp(K)的简化背景SCimp(K)=(Simp(K2),K3,RSimp(K2)),其中,

表4   简化背景SCimp(K)

Table 4  The reduced context SCimp(K)

c1c2c3
m1001
m2010
m1m3100
m3m1101

新窗口打开| 下载CSV


Simp(K2)=              m1,m2,m1m3,m3m1

对应的概念格L(SCimp(K))图2所示.

图2

图2   概念格L(SCimp(K))

Fig.2   The concept lattice L(SCimp(K))


对比图1图2可知,概念格L(Cimp(K))L(SCimp(K))同构.

3 基于条件属性蕴含的对象定向概念格

在条件属性蕴含背景Cimp(K)上引入必然性算子和可能性算子,:2imp(K2)2K3:

:= {cK3Amimp(K2),
(Am,c)RAm}
:= {cK3Amimp(K2),
(Am,c)RAm}

类似地定义另外一对近似算子:

,:2K32imp(K2)
C:= {Amimp(K2),cK3,
(Am,c)RcC}
C:= {Amimp(K2),cK3,
(Am,c)RcC}

定义11K=(K1,K2,K3,Y)为三元背景,Cimp(K)=(imp(K2),K3,R)为基于K的条件属性蕴含背景,imp(K2),CK3.若二元组(,C)满足:=C,C=,则称(,C)为条件属性蕴含背景Cimp(K)的对象定向概念,并分别称C为该对象定向概念(,C)的外延和内涵.

对于任意的两个对象定向概念(1,C1)(2,C2)定义上、下确界:

(1,C1)(2,C2)=((C1C2),C1C2)
(1,C1)(2,C2)=(12,(12))

记:

LO(Cimp(K))={(,C)|=C,C=}

(LO(Cimp(K)),,)构成一个完备格,称为条件属性蕴含背景Cimp(K)的对象定向概念格.

例4 计算表3中条件属性蕴含背景Cimp(K)的八个对象定向概念如下:

O1:=({c1},{c1})=({m1m3,m3m1,{m1,m2}m3,{m2,m3}m1},{c1})O2:=({c2},{c2})=({m2,m1m2,m3m2,{m1,m3}m2},{c2})
O3:=({c3},{c3})=({m1,m2m1,m3m1,{m2,m3}m1},{c3})O4:=({c1},{c1})({c2},{c2})=(m2,m1m2,m1m3,m3m1,m3m2,{m1,m2}m3,{m1,m3}m2,{m2,m3}m1,{c1,c2})
O5:=({c1},{c1})({c3},{c3})=(m1,m1m3,m2m1,m3m1,{m1,m2}m3,{m2,m3}m1,{c1,c3})O6:=({c2},{c2})({c3},{c3})=({m1,m2,m1m2,m2m1,m3m1,m3m2,{m1,m3}m2,{m2,m3}m1},{c2,c3})O7:=({c1},{c1})({c2},{c2})({c3},{c3})=(imp(K2),K3)O8:=({c1},{c1})({c2},{c2})({c3},{c3})=(,)

图3给出了表3条件属性蕴含背景Cimp(K)的对象定向概念格LO(Cimp(K)).

图3

图3   对象定向概念格LO(Cimp(K))

Fig.3   The object oriented concept lattice LO(Cimp(K))


定理2K=(K1,K2,K3,Y)为三元背景,Cimp(K)为条件属性蕴含背景,SCimp(K)Cimp(K)的简化背景,则对象定向概念格LO(Cimp(K))LO(SCimp(K))同构.

定理2的证明类似定理1,不再具体证明.

例5(续例4) 计算表4所示简化背景SCimp(K)对应的对象定向概念格LO(SCimp(K)),如图4所示.其中,

图4

图4   对象定向概念格LO(SCimp(K))

Fig.4   The object oriented concept lattice LO(SCimp(K))


SO1:=({m1m3,m3m1},{c1})
SO2:=({m2},{c2})
SO3:=({m1,m3m1},{c3})
SO4:=({m2,m1m3,m3m1},{c1,c2})
SO5:=({m1,m1m3,m3m1},{c1,c3})
SO6:=({m1,m2,m3m1},{c2,c3})
SO7:=(Simp(K2),K3)
SO8:=(,)

对比图3图4可知,对象定向概念格LO(Cimp(K))LO(SCimp(K))同构.

4 基于条件属性蕴含的属性定向概念格

定义12K=(K1,K2,K3,Y)为三元背景,Cimp(K)=(imp(K2),K3,R)为基于K的条件属性蕴含背景,imp(K2),CK3.若二元组(,C)满足=C,C=,则称(,C)为条件属性蕴含背景Cimp(K)的属性定向概念,并分别称C为概念(,C)的外延和内涵.

对于任意的两个属性定向概念(1,C1)(2,C2)定义上、下确界:

(1,C1)(2,C2)=(12,(12))
(1,C1)(2,C2)=((C1C2),C1C2)

LP(Cimp(K))={(,C)|=C,C=},则(LP(Cimp(K)),,)构成一个完备格,称其为条件属性蕴含背景Cimp(K)的属性定向概念格.

例6 计算表3中条件属性蕴含背景Cimp(K)对应的八个属性定向概念如下:

P1:=({c1},{c1})=({m1m3,{m1,m2}m3},{c1})
P2:=({c2},{c2})=({m2,m1m2,m3m2,{m1,m3}m2},{c2})
P3:=({c3},{c3})=({m1,m2m1},{c3})P4:=({c1},{c1})({c2},{c2})=({m2,m1m2,m1m3,m3m2,m3m1,{m1,m2}m3,{m1,m3}m2,{m2,m3}m1},{c1,c2})
P5:=({c1},{c1})({c3},{c3})=({m1,m1m3,m2m1,m3m1,{m1,m2}m3,{m2,m3}m1,{c1,c3})P6:=({c2},{c2})({c3},{c3})=({m1,m2,m1m2,m2m1,m3m2,{m1,m3}m2,{c2,c3})P7:=({c1},{c1})({c2},{c2})({c3},{c3})=(imp(K2),K3)P8:=({c1},{c1})({c2},{c2})({c3},{c3})=,

图5给出了表3条件属性蕴含背景Cimp(K)的属性定向概念格LP(Cimp(K)).

图5

图5   属性定向概念格LP(Cimp(K))

Fig.5   The property oriented concept lattice LP(Cimp(K))


同样的,简化背景SCimp(K)与原背景Cimp(K)上的属性定向概念格也有如下的关系:

定理3K=(K1,K2,K3,Y)为三元背景,Cimp(K)为条件属性蕴含背景,SCimp(K)Cimp(K)简化背景,则属性定向概念格LP(Cimp(K))LP(SCimp(K))同构.

例7(续例6) 考虑表4所示简化背景SCimp(K)对应的属性定向概念,计算得:

SP1:=({m1m3},{c1})
SP2:=({m2},{c2})
SP3:=({m1},{c3})
SP4:=({m2,m1m3,m3m1},{c1,c2})
SP5:=({m1,m1m3,m3m1},{c1,c3})
SP6:=({m1,m2},{c2,c3})
SP7:=(Simp(K2),K3)
SP8:=(,)

表4   简化背景SCimp(K)对应的属性定向概念格LP(SCimp(K))如图6所示.

图6 属性定向概念格LP(SCimp(K))

Fig.6 The property oriented concept lattice LP(SCimp(K))

新窗口打开| 下载CSV


同样的,由图5和图6可知,属性定向概念格LPCimp(K)LPSCimp(K)同构.

5 总 结

本文利用三元背景的条件属性蕴含构造了条件属性蕴含形式背景,并给出了形式概念的定义.由于实际问题中条件属性蕴含形式背景往往是一个比较大的数据表,因此本文在保持概念格同构的原则下对条件属性蕴含形式背景的对象进行约简.最后,给出了条件属性蕴含形式背景的对象定向概念格和属性定向概念格的定义及对象约简方法.

参考文献

LehmannFWilleR.

A triadic approach to formal concept analysis∥Ellis G,Levinson R,Rich W,

et al. Conceptual Structuresapplications,Implemen⁃tation and Theory. Springer Berlin Heidelberg199532-43.

[本文引用: 2]

WilleR.

The basic theorem of triadic concept analysis

Order,199512(2):149-158.

[本文引用: 1]

WilleR.

Restructuring lattice theory:an approach based on hierarchies of concepts∥Rival I

NATO Advanced Study Institute (Ordered Sets). Springer Berlin Heidelberg,1982445-470.

[本文引用: 4]

GanterBWilleR.

Formal concept analysis:mathematical foundations

Springer Berlin Heidelberg,1999,17-61.

[本文引用: 2]

WeiLQianTWanQet al.

A research summary about triadic concept analysis

International Journal of Machine Learning and Cybernetics,20189(4):699-712.

[本文引用: 1]

魏玲,万青,钱婷.

三元概念分析综述

.西北大学学报(自然科学版),2014,44(5):689-699.

[本文引用: 1]

Wei L,Wan Q,Qian T,et al.

An overview of triadic concept analysis.

Journal of Northwest University (Natural Science Edition)201444(5):689-699.

[本文引用: 1]

StummeG.

Attribute exploration with background implications and exceptions∥Bock H H,Polasek W

Data Analysis and Information Systems. Springer Berlin Heidelberg,1996457-469.

[本文引用: 1]

BělohlávekRVychodilV.

Attribute implications in a fuzzy setting∥Missaoui R,Schmid J

Formal Concept Analysis. Springer Berlin Heidelberg,200645-60.

[本文引用: 1]

VychodilV.

On the importance of fuzzy attribute implications∥2008 IEEE International Conference on Fuzzy Systems

Hong Kong,ChinaIEEE2008281-286.

[本文引用: 1]

VychodilV.

On minimal sets of graded attribute implications

Information Sciences,2015294478-488.

[本文引用: 1]

ZhaiY HLiD YQuK S.

Decision implication canonical basis:a logical perspective

Journal of Computer and System Sciences,201581(1):208-218.

[本文引用: 1]

Rodríguez⁃LorenzoEAdarichevaKCorderoPet al.

From an implicational system to its corresponding D⁃basis∥Sadok B Y,Jan K

The 12th International Conference on Concept Lattices and Their Applications. Clermont⁃Ferrand,FranceBlaise Pascal University2015217-228.

[本文引用: 1]

安秋生孔祥玉.

函数依赖与属性蕴含的关系研究

小型微型计算机系统,201738(9):2000-2005.

[本文引用: 1]

An Q S,Kong X Y.

Relationship study of functional dependency and attribute implication.

Journal of Chinese Computer Systems201738(9):2000-2005.

[本文引用: 1]

TriskaJVychodilV.

Logic of temporal attribute implications

Annals of Mathematics and Artificial Intelligence,201779(4):307-335.

[本文引用: 1]

VityaevE EDeminA VPonomaryovD K.

Probabilistic generalization of formal concepts

Programming and Computer Software,2012

[本文引用: 1]

38(5):219-230.

[本文引用: 1]

BiedermannK.

A foudation of the theory of trilatttices

Ph. D.Dissertation. Aachen:TU Darmstadt1998.

[本文引用: 1]

GanterBObiedkovS A.

Implications in triadic formal contexts∥Wolff K E,Pfeiffer H D,Delugach H S

Conceptual Structures. Springer Berlin Heidelberg,2004186-195.

[本文引用: 3]

GlodeanuC V.

Fuzzy⁃valued triadic implications∥Napoli A,Vychodil V

The 8th International Conference on Concept Lattices and Their Applications. Springer Berlin Heidelberg,2011159-173.

[本文引用: 1]

Rodríguez⁃LorenzoECorderoPEncisoMet al.

CAISL:simplification logic for conditional attribute implications∥Huchard M,Kuznetsov S O

The 13th International Conference on Concept Lattices and Their Applications. Moscow,RussiaNational Research University2016337-348.

[本文引用: 1]

Rodríguez⁃LorenzoECorderoPEncisoMet al.

An axiomatic system for conditional attribute implications in triadic concept analysis

International Journal of Intelligent Systems,201732(8):760-777.

[本文引用: 1]

TangY QFanMLiJ H.

An information fusion technology for triadic decision contexts

International Journal of Machine Learning and Cybernetics,20167(1):13-24.

[本文引用: 1]

祁建军魏玲.

三元背景及概念三元格的简化

计算机科学,201744(9):53-57.

[本文引用: 1]

Qi J J,Wei L.

Simplification of triadic contexts and concept trilattices.

Computer Science201744(9):53-57.

[本文引用: 1]

王霞张茜李俊余.

基于粗糙集的三元概念分析

山东大学学报(理学版),201752(7):37-43.

[本文引用: 1]

Wang X,Zhang Q,Li J Y,et al.

Triadic concept analysis based on rough set theory.

Journal of Shandong University (Natural Science)201752(7):37-43.

[本文引用: 1]

李俊余朱荣杰王霞.

三元概念与形式概念的关系

南京大学学报(自然科学),201854(4):786-793.

[本文引用: 1]

Li J Y,Zhu R J,Wang X,et al.

The relationship between triadic concepts and formal concepts.

Journal of Nanjing University (Natural Science)201854(4):786-793.

[本文引用: 1]

王霞江山李俊余.

三元概念的一种构造方法

计算机研究与发展,201956(4):544-853.

[本文引用: 1]

Wang X,Jiang S,Li J Y,et al.

A construction method of triadic concepts.

Journal of Computer Research and Development201956(4):544-853.

[本文引用: 1]

HwangS H.

A triadic approach of hierarchical classes analysis on folksonomy mining

International Journal of Computer Science & Network Security,20077(8):193-198.

[本文引用: 1]

HwangS HKimE H.

An approach for mining folksonomies based on the triadic class hierarchies

Advanced Science Letters,20129(1):844-849.

[本文引用: 1]

刘晓今.

概念三元格构造算法及应用研究

博士学位论文. 西安西安电子科技大学2013.

[本文引用: 1]

Liu X J.

Study on the construction algorithem of concept trilattices and its application

Ph. DDissertation. Xi’an:Xidian University2013.

[本文引用: 1]

王寿彪李新明刘东.

基于大数据的装备体系三元概念认知系统模型构造框架

系统工程与电子技术,201638(11):2537-2545.

[本文引用: 1]

Wang S B,Li X M,Liu D,et al.

Constructing framework of triadic concepts cognitive system model of equipment system of systems based on big data.

Systems Engineering and Electronics201638(11):2537-2545.

[本文引用: 1]

李贞张卓王黎明.

基于三元概念分析的文本分类算法研究

计算机科学,201744(8):207-215.

[本文引用: 1]

Li Z,Zhang Z,Wang L M.

Research on text classification algorithm based on triadic concept analysis.

Computer Science201744(8):207-215.

[本文引用: 1]

王红敏张卓王黎明.

三元概念分析的联盟应用研究

小型微型计算机系统,201839(12):2571-2576.

[本文引用: 1]

Wang H M,Zhang Z,Wang L M.

Application research on coalition of triadic concept analysis.

Journal of Chinese Computer Systems201839(12):2571-2576.

[本文引用: 1]

HaoFParkD SMinG Yet al.

K⁃Cliques mining in dynamic social networks based on triadic formal concept analysis

Neurocomputing,201620957-66.

[本文引用: 1]

KumarC AMouliswaranS CLiJ Het al.

Role based access control design using triadic concept analysis

Journal of Central South University,201623(12):3183-3191.

[本文引用: 1]

/