Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Save to My Library
Look up keyword
Like this
1Activity
0 of .
Results for:
No results containing your search query
P. 1
Mining Rules from Crisp Attributes by Rough Sets on the Fuzzy Class Sets

Mining Rules from Crisp Attributes by Rough Sets on the Fuzzy Class Sets

Ratings:

5.0

(1)
|Views: 48 |Likes:
Published by ijcsis
Machine learning can extract desired knowledge and ease the development bottleneck in building expert systems. Among the proposed approaches, deriving classification rules from training examples is the most common. Given a set of examples, a learning program tries to induce rules that describe each class. The rough-set theory has served as a good mathematical tool for dealing with data classification problems. In the past, the rough-set theory was widely used in dealing with data classification problems that data sets were containing crisp attributes and crisp class sets. This paper thus extends rough-set theory previous approach to deal with the problem of producing a set of certain and possible rules from crisp attribute by rough sets on the fuzzy class sets. The proposed approach combines the rough-set theory and the fuzzy class sets theory to learn. The examples and the approximations then interact on each other to drive certain and possible rules. The rules derived can then serve as knowledge concerning the data sets on the fuzzy class sets.
Machine learning can extract desired knowledge and ease the development bottleneck in building expert systems. Among the proposed approaches, deriving classification rules from training examples is the most common. Given a set of examples, a learning program tries to induce rules that describe each class. The rough-set theory has served as a good mathematical tool for dealing with data classification problems. In the past, the rough-set theory was widely used in dealing with data classification problems that data sets were containing crisp attributes and crisp class sets. This paper thus extends rough-set theory previous approach to deal with the problem of producing a set of certain and possible rules from crisp attribute by rough sets on the fuzzy class sets. The proposed approach combines the rough-set theory and the fuzzy class sets theory to learn. The examples and the approximations then interact on each other to drive certain and possible rules. The rules derived can then serve as knowledge concerning the data sets on the fuzzy class sets.

More info:

Published by: ijcsis on May 17, 2012
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

06/05/2012

pdf

text

original

 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 10, No. 4, April 2012
Mining Rules from Crisp Attributes by RoughSets on the Fuzzy Class Sets
Mojtaba MadadyarAdeh
#1
, Dariush Dashchi Rezaee
#2
, Ali Soultanmohammadi
#3
#
Sama Technical and Vocational Training College, Islamic Azad University, Urmia BranchUrmia, Iran
1
m.madadyar@iaurmia.ac.ir
2
d_dashchi_rezaee@yahoo.com
3
ali_soultanmohammadi@yahoo.com
 
 Abstract
 — 
Machine learning can extract desiredknowledge and ease the development bottleneck inbuilding expert systems. Among the proposedapproaches, deriving classification rules from trainingexamples is the most common. Given a set of examples,a learning program tries to induce rules that describeeach class. The rough-set theory has served as a goodmathematical tool for dealing with data classificationproblems. In the past, the rough-set theory was widelyused in dealing with data classification problems thatdata sets were containing crisp attributes and crisp classsets. This paper thus extends rough-set theory previousapproach to deal with the problem of producing a set of certain and possible rules from crisp attribute by roughsets on the fuzzy class sets. The proposed approachcombines the rough-set theory and the fuzzy class setstheory to learn. The examples and the approximationsthen interact on each other to drive certain and possiblerules. The rules derived can then serve as knowledgeconcerning the data sets on the fuzzy class sets.
 Keywords-Fuzzy set; Rough set; Data mining; Fuzzy class sets; Crisp attributes; Certain rule; Possibl 
e rule; α
- cut
I.
 
I
NTRODUCTION
 Machine learning and data mining techniques haverecently been developed to find implicitly meaningfulpatterns and ease the knowledge-acquisitionbottleneck. Among these approaches, derivinginference or association rules from training examplesis the most common [11], [13]. Given a set of examples and counterexamples of a concept, thelearning program tries to induce general rules thatdescribe all or most of the positive training instancesand none or few of the counterexamples [6]. If thetraining instances belong to more than two classes, thelearning program tries to induce general rules thatdescribe each class. Recently, the rough-set theory hasbeen used in reasoning and knowledge acquisition forexpert systems [3][13]. It was proposed by Pawlak in1982, with the concept of equivalence classes as itsbasic principle. Several applications and extensions of the rough-set theory have also been proposed.
Examples are Orlowska‘s reasoning with incomplete
information, [1] knowledge-base reduction, [9] datamining, Zhong, Dong, [18] rule discovery. Due to thesuccess of the rough-set theory to knowledgeacquisition, many researchers in database andmachine learning fields are interested in this newresearch topic because it offers opportunities todiscover useful information in training examples. [19]Mentioned that the main issue in the rough-setapproach was the formation of good rules. Hecompared the rough-set approach with some otherclassification approaches .The main characteristic of the rough-set approach lies in that it can use the notionof inadequacy of available information to performclassification of objects [19][20]. It can also form anapproximation space for analysis of informationsystems. Partial classification may be formed from thegiven objects. Ziarko also mentioned the limitations of the rough-set model. For example, the classificationwith a controlled degree of uncertainty ormisclassification error is outside the realm of theapproach. Overgeneralization is another limitation tothe rough-set approach. Ziarko thus proposed thevariable precision rough-set model to solve the aboveproblems .The variable precision rough-set model hashowever only shown how binary or crisp valuedtraining data may be handled. Training data in real-world applications usually consist of quantitativevalues. Although the variable precision rough-setmodel can also manage the quantitative values bytaking each quantitative value as an attribute value,the rules formed in this way may be too specific. Itmay also cause humans hard to interpret them.Extending the variable precision rough-set model toeffectively dealing with quantitative values is thusimportant to real applications of the model. Since thefuzzy set concepts are often used to representquantitative data by linguistic terms and membershipfunctions because of their simplicity and similarity tohuman reasoning [2], we thus attempt to combine thevariable precision rough-set model and the fuzzy settheory to solve the above problems. The rules minedare expressed in linguistic terms, which are more
112http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 10, No. 4, April 2012
natural and understandable for human beings. Sincethe number of linguistic terms is much less than thatof possible quantitative values, the over-specializationproblem can be avoided. Tzung [7] has successfullyproposed a mining algorithm to find fuzzy rules basedon the rough-set model. The variable precision rough-set model can be thought of as a generalization of therough-set model. Tzung [10] deal whit the problem of producing a set of certain and possible rules fromincomplete data sets on the crisp class sets.In this paper, we thus deals with the problem of producing a set of certain and possible rules frommining crisp attributes by rough sets on the fuzzyclass sets . A new method, approach combines therough-set theory and the fuzzy class sets theory tolearn, is thus proposed to solve this problem. It firsttransforms each class sets quantitative value into afuzzy set of linguistic terms using membershipfunctions and converts
each of fuzzy class sets by α
-cut in several crisp subclasses. It second, calculatesthe lower and the upper approximations. The certainand possible rules are then generated based on theseapproximations. This paper thus extends rough-settheory previous approach to deal with the problem of producing a set of certain and a possible rule fromcrisp attributes by rough sets on the fuzzy class sets.The paper thus extends the existing rough-set miningapproaches to process quantitative data with toleranceof noise and uncertainty.The remaining parts of this paper are organized asfollows. In Section 2, the variable precision rough-setmodel is reviewed. In Section 3,
α
-cut and fuzzy classsets is described. In Section 4, the notation used inthis paper is described. In Section 5, the proposedalgorithm for crisp attributes data sets on the fuzzyclass sets. In Section 6, an example is given toillustrate the proposed algorithm.II.
 
R
EVIEW OF THE ROUGH
-
SET THEORY
 The rough-set theory, proposed by Pawlak in 1982[14], can serve as a new mathematical tool for dealingwith data classification problems. It adopts theconcept of equivalence classes to partition traininginstances according to some criteria. Two kinds of partitions are formed in the mining process: lowerapproximations and upper approximations, fromwhich certain and possible rules can easily be derived.Formally, let U be a set of training examples (objects),A be a set of attributes describing the examples, C bea set of classes, and V
 j
be a value domain of anattribute A
 j
. Also let v
 j(i)
be the value of attribute A
 j
 for the ith object Obj
(i)
. When two objects Obj
(i)
andObj
(k)
have the same value of attribute A
 j
, (that is, v
 j(i)
 = v
 j(k)
), Obj
(i)
and Obj
(k)
are said to have anindiscernibility relation (or an equivalence relation) onattribute A
 j
. Also, if Obj
(i)
and Obj
(k)
have the samevalues for each attribute in subset B of A; Obj
(i)
andObj
(k)
are also said to have an indiscernibility(equivalence) relation on attribute set B. Theseequivalence relations thus partition the object set Uinto disjoint sub sets, denoted by U/B, and thepartition including Obj
(i)
is denoted by B(Obj
(i)
). Theset of equivalence classes for subset B is referred to asB-elementary set.
Example 1
. Table I shows a data set containingseven objects denoted by U ={ Obj
(1)
; Obj
(2)
;...; Obj
(7)
 }, two attributes denoted by A={Systolic Pressure(SP), Diastolic Pressure (DP)}, and a class set BloodPressure (BP). Assume the attributes and the classesset have three possible values: {Low (L), Normal (N)and High (H)}.
 
TABLE I. T
HE DATA SET FOR EXAMPLE
1.
Object Systolic Pressure(SP) Diastoli Pressure(DP) Blood  Pressure(BP)
 obj
(1)
 
L N L
 obj
(2)
 
H N H
 obj
(3)
 
N N N
 obj
(4)
 
L L L
 obj
(5)
 
H H H
 obj
(6)
 
N H H
 obj
(7)
 
N L N
Since Obj
(1)
and Obj
(4)
have the same attributevalue (L) for attribute SP, they share anindiscernibility relation and thus belong to the sameequivalence class for SP. The equivalence partitions(elementary sets) for singleton attributes can bederived as follows:U/{SP} = {{obj
(2)
, obj
(5)
}{ obj
(3)
, obj
(6)
, obj
(7)
}{ obj
(1)
,obj
(4)
}}, andU/{DP} = {{obj
(1)
, obj
(2)
, obj
(3)
}{ obj
(4)
, obj
(7)
}{obj
(5)
, obj
(6)
}},Also, {SP}( obj
(1)
) = {SP}( obj
(4)
) = { obj
(1)
, obj
(4)
}.The rough-set approach analyzes data according totwo basic concepts, namely the lower and the upperapproximations of a set. Let X is an arbitrary subset of the universe U, and B is an arbitrary subset of attributeset A. The lower and the upper approximations for Bon X denoted B
*
(X) and B
*
(X) respectively, aredefined as follows [20] [4]:B
*
(X) = {x|x
ϵ
U, B(X)
X}(1)B
*
(X) = {x|x
ϵ
 
U and B(X) ∩ X ≠ Ø}
(2)Elements in B
*
(X) can be classified as members of set X with full certainty using attribute set B, so B
*
(X)is called the lower approximation of X. Similarly,elements in B
*
(X) can be classified as members of theset X with only partial certainty using attribute set B,so B
*
(X) is called the upper approximation of X.
113http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 10, No. 4, April 2012
Example2
. Continuing from Example 1, assumeX={Obj
(1)
,Obj
(4)
}. The lower and the upperapproximations of attribute DP with respect to X canbe calculated as follows:DP
*
(X) = Ø, andDP
*
(X) = {{ obj
(1)
, obj
(2)
, obj
(3)
}{ obj
(4)
, obj
(7)
}}.After the lower and the upper approximations havebeen found, the rough-set theory can then be used toderive certain information and induce certain andpossible rules from them (Grzymala-Busse, 1988).III.
 
Α
-
CUT AND FUZZY CLASS SETS
 
An α
-level set of a fuzzy set A of X is a non-fuzzydenoted by [A]
α
and is defined by,
0))(supp( 0)(|{ [A]
   
 
if  Aclif  A X 
 (3)Where cl (supp(A)) denotes the closure of thesupport of A.
Definition 1
(Support) Let A be a fuzzy subset of X; the support of A, denoted supp(A), is the crispsubset of X whose element all have nonzeromembership grades in A.
}.0)(|{)(sup
x A X  x A p
 (4)
Definition 2
(triangular fuzzy number) A fuzzy setA is called triangular fuzzy number with peak (or
center) a, left width α>0 and right width β>0 if its
membership function has the following from,
otherwiseaifaa aifaa
0 / )(1  / )(1 A(t)
     
 (5)And we use the notation A= (a
, α, β
). It can easilybe verified that,
].1,0[],)1(,)1([[A]
      
 
aa
 (6)The support of A is (a-
α
, a
+β).
In the past, the rough-set theory was widely used in dealing with dataclassification problems [10]. Most conventionalmining algorithms based on the rough-set theoryidentify relationships among data using crisp classsets values. This possible exist class sets withquantitative values, however, are commonly seen inreal-world applications. In this paper, we thus dealwith the problem of learning from class quantitativedata sets based on rough sets. A learning algorithm isproposed, which can simultaneously derive certainand possible rules from class quantitative data sets.Class sets with quantitative values are firsttransformed into fuzzy sets of linguistic terms usingmembership functions. Therefore, convert fuzzy class
sets with α
-cut define to several crisp subclasses.Number of divisions arbitrary,
that α
-cut perform onthe linguistic terms.IV.
 
N
OTATION
 Notation used in this paper is described as follows:
U
universe of all objects
n
total number of training examples(objects) in U
Obj
(i)
 
ith training example (object), 1 ≤i ≤n
 
A
set of all attributes describing U
m
total number of attributes in A
B
an arbitrary subset of A
A
 j
 
 jth attribute, 1≤ j≤ m
 |
A
 j
|
number of attribute values for Aj
v
 j(i)
 
the value of A
 j
for Obj
(i)
 
d
number of divisions arbitrary , that
α
-cutperform on the linguistic terms
C
set of classes to be determined
c
total number of classes in C
R
k
 
kth fuzzy region of C,1 ≤k ≤c
 
e
(i)
 
the value of C for Obj
(i)
(i)
the fuzzy set converted from e
(i)
 
k(i)
the membership value of e
(i)
in region R
 
X
l
lth class, 1 ≤ l≤ (c×d)
 
B(Obj
(i)
)
the fuzzy incomplete equivalenceclasses in which Obj
(i)
exists
B
*
(X)
the fuzzy incomplete lower approximationfor B on X
B
*
(X)
the fuzzy incomplete upper approximationfor B on XThese fuzzy equivalence relations thus partitionthe fuzzy object set U into several fuzzy subsets thatmay overlap, and the result is denoted by U/B. The setof partitions, based on B and including Obj
(i)
, isdenoted B(Obj
(i)
). Thus, B(Obj
(i)
)= {(B
1
(Obj
(i)
)
(B
r
(Obj
(r)
) }, where r is the number of partitionsincluded in B(Obj
(i)
).
114http://sites.google.com/site/ijcsis/ISSN 1947-5500

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->