You are on page 1of 6

INTERNATIONAL JOURNAL OF COMPUTATIONAL COGNITION (HTTP://WWW.IJCC.US), VOL. 6, NO.

4, DECEMBER 2008 45
A Study of Fuzzy Relational Database
Jaydev Mishra and Sharmistha Debnath Ghosh
AbstractThe present work deals with detail study of func-
tional dependency in fuzzy relational databases. The paper is
an extension of the work done by Al-Hamouz and Biswas [1].
Here, the f-Armstrongs axioms of [1] have been modied with
verication and the well-known concepts of partial ffd, fuzzy
key, fuzzy closure of attribute(s) and redundant ffd have been
studied with the new notion of -ffd as dened in [1]. Finally
fuzzy key and fuzzy closure of attribute(s) have been tested with
a real life example. Copyright c 2008 Yangs Scientic Research
Institute, LLC. All rights reserved.
Index TermsFuzzy set, fuzzy union, fuzzy functional de-
pendency(ffd), f-Armstrongs axioms, partial ffd, fuzzy key,
redundant ffd.
I. INTRODUCTION
I
N recent years, a lot of attention has been attracted to fuzzy
relational databases that generalize the classical relational
data model introduced by Codd [5] by allowing uncertain and
imprecise information to be represented and manipulated. Data
is often partially known, vague or ambiguous in many real
world applications. Fuzziness is introduced in the classical
model to deal with such imprecise information and several ex-
tensions of the model are available in literature [1][3][4][6][7].
Data dependency plays a crucial role in logical database
design and functional dependency of one set of attributes upon
another is one of the most important concepts in relational
databases. When data is of fuzzy nature, the concept of
functional dependency has been extended by several authors
[1][2][4][7][8] to dene fuzzy functional dependency(ffd) us-
ing the concept of fuzzy logic. The comparison of two data
of a domain is done with the help of fuzzy equality relations
in these existing concepts of ffd. Recently, in 2006, a new
notion of ffd based on equivalence relation has been introduced
by Al-Hamouz and Biswas [1] and the f-Armstrongs axioms
have been studied and veried with that ffd. The denition of
ffd dened in [1] with the concept of (

)-nearer and (

)-
equality made easy for developing the theory of fuzzy database
compare to the denition of ffd dened by other authors.
The present work is based on the ffd dened by Al-Hamouz
and Biswas which introduces a choice parameter [0, 1]
that may be set by the database designer and is called -
ffd. In this paper, the f-Transitive, f-Union, f-Pseudotransitive
Manuscript received May 13, 2008; revised July 09, 2008.
Jaydev Mishra, Department of CSE & IT, College of Engi-
neering and Management, Kolaghat-721171, West Bengal, India.
Email: jsm02@rediffmail.com. Sharmistha Debnath Ghosh, Department
of Mathematics, College of Engineering and Management, Kolaghat-721171,
West Bengal, India. Email: sharmisthag@yahoo.com.
Acknowledgment: The author would like to thank Prof. Ranjit Biswas for
their contribution to the eld of fuzzy databases.
Publisher Item Identier S 1542-5908(08)10403-1/$20.00
Copyright c 2008 Yangs Scientic Research Institute, LLC. All
rights reserved. The online version posted on February 17, 2009 at
http://www.YangSky.com/ijcc/ijcc64.htm
and f-Decomposition rules of [1] have been modied with
proper verication. I have also extended the work to dene
partial fuzzy functional dependency(partial ffd), fuzzy key,
fuzzy closure of an attribute or a set of attributes, redundant
fuzzy functional dependency(redundant ffd).
The paper is organized as follows: In section II, we have
revisited some basic denitions of fuzzy set theory and then
dened fuzzy functional dependency(ffd) as in [1]. In section
III, the f-Armstrongs axioms have been studied and modied
with verication. Next, in section IV, we have proceeded to
dene partial fuzzy functional dependency(partial ffd). The
denition of fuzzy key at -level of choice has been introduced
and discussed with an example in section V. The idea of
fuzzy closure of an attribute or a set of attributes has been
studied in section VI together with an algorithm and an
example. An example of real life application area has been
taken in section VII to show how fuzzy key and fuzzy closure
of attribute(s) can be computed. Redundant fuzzy functional
dependency(redundant ffd) has been dened in section VIII
with example. Finally, the concluding remarks have been given
in the last section IX.
II. FUZZY FUNCTIONAL DEPENDENCY(FFD)
In this section, I rst review some basic denitions of fuzzy
set theory that will be useful throughout the paper and then
dene the Fuzzy Functional Dependency ffd) as introduced in
[1].
Let U = {u
1
, u
2
, . . . , u
n
} be a universe of discourse.
Denition II.1 A fuzzy set A in the universe of discourse U
is characterized by the membership function
A
given by
A
:
U [0, 1] and A is dened as the set of ordered pairs
A = {(u,
A
(u)) : u U}, where
A
(u) is the grade of
membership of element u in the set A.
Denition II.2 If A and B are two fuzzy sets of the universe
U, then the fuzzy union of A and B is denoted by A

fuzzy
B
and is dened as A

fuzzy
B = {(x, max{
A
(x),
B
(x)}) :
x U}.
Denition II.3 Let X and Y be two sets. A fuzzy relation
R from X to Y is a fuzzy set on X Y and is denoted by
R(X Y ).
Denition II.4 A fuzzy relation R(X X) is said to be
1) reexive: iff x X,
R
(x, x) = 1;
2) symmetric: iff x
1
, x
2
X,
R
(x
1
, x
2
) =
R
(x
2
, x
1
).
A fuzzy relation is said to be a fuzzy tolerance relation if it is
reexive and symmetric.
46 INTERNATIONAL JOURNAL OF COMPUTATIONAL COGNITION (HTTP://WWW.IJCC.US), VOL. 6, NO. 4, DECEMBER 2008
Next, to introduce the new notion of ffd as dened in [1],
we give the following denitions and terminologies.
Let X be a universal set and be a fuzzy tolerance relation
on X. Consider a choice parameter [0, 1] to be predened
by the database designer.
Denition II.5 ()

-nearer or -nearer elements. Two ele-


ments x
1
, x
2
X are said to be ()

-nearer (or -nearer),


if

(x
1
, x
2
) , we denote this by the notation x
1
N
()

x
2
.
Denition II.6 ()

-equality or-equality elements. Two el-


ements x
1
, x
2
X are said to be ()

-equal(or -equal)
if
1) either x
1
N
(
)
x
2
or
2) y
1
, y
2
, y
3
, . . . , y
r1
, y
r
X such that
{x
1
N
()

y
1
, y
1
N
()

y
2
, . . . , y
r1
N
()

y
r
, y
r
N
()

x
2
}.
This equality is denoted by the notation x
1
E
()

x
2
.
Denition II.7
()

relation on X. The crisp relation


()

on X is dened as: For x


1
, x
2
X, x
1

()

x
2
if x
1
E
()

x
2
.
Proposition II.1 The relation
()

dened on X is an equiv-
alence relation (See ref. [1])
Now, consider a relation r(R) of a relation schema
R(A
1
, A
2
, . . . , A
n
). Let us assume a fuzzy tolerance relation

i
on the domain dom(A
i
), i = 1, 2, . . . , n. Let denote
the set {
1
,
2
, . . . ,
n
} of fuzzy tolerance relations. Let
X = {X
1
, X
2
, . . . , X
k
} R. We now dene ()

-equality
of two tuples t
1
[X] and t
2
[X] in a relational database design.
Denition II.8 ()

-equality of t
1
[X] and t
2
[X]. Two tu-
ples t
1
[X] and t
2
[X] are said to be ()

-equal if
t
1
[X
i
]E
()

i
t
2
[X
i
], i = 1, 2, . . . , k. The equality notation
is denoted as t
1
[X]
()

t
2
[X] or simply by t
1
[X]

t
2
[X].
The following results are then obtained straight forward.
Proposition II.2 1) For any tuple t and for any [0, 1],
t[X]
()

t[X];
2) t
1
[X]
()

t
2
[X] t
2
[X]
()

t
1
[X];
3) If 0
2

1
1, then t
1
[X]
1
t
2
[X]
t
1
[X]
2
t
2
[X].
Denition II.9 Fuzzy Functional Dependency (ffd): Let
X, Y R = {A
1
, A
2
, . . . , A
n
}. Choose a parameter
[0, 1] and propose a fuzzy tolerance relation . A fuzzy
functional dependency(ffd), denoted by X
()

Y or simply
by X

Y , is said to exist, if whenever t


1
[X]
()

t
2
[X], it
is also the case that t
1
[Y ]
()

t
2
[Y ].
This ffd can be read as X fuzzy functionally determines
Y at -level of choice or Y fuzzy functionally depends on
X at -level of choice and is called an -ffd. Clearly, by
denition of -ffd, it follows that for any subset X of R and
for any [0, 1], X

X. Further we have the following


propositions:
Proposition II.3 If 0
2

1
1, then X

1
Y
X
2
Y .
Proof: Let us suppose that the ffd X

1
Y holds. Then,
by denition of ffd, we have
t
1
[X]

1
t
2
[X] t
1
[Y ]

1
t
2
[Y ] (1)
Now by Proposition II.2 3), we have for 0
2

1
1,
t
1
[Y ]

1
t
2
[Y ] t
1
[Y ]

2
t
2
[Y ].
Using above relations in Eq. (1), we can rewrite Eq. (1) as
t
1
[X]

1
t
2
[X] t
1
[Y ]

2
t
2
[Y ] (2)
Now Eq. (2) states that whenever t
1
[X] and t
2
[X] are
1
-equal
then it is the case that t
1
[Y ] and t
2
[Y ] are
2
-equal. Again
by Proposition II.2 3), we can say, if t
1
[X] and t
2
[X] are
1
-
equal then t
1
[X] and t
2
[X] are also
2
-equal (
2

1
).
So, nally we can state that t
1
[X]

2
t
2
[X]
t
1
[Y ]
2
t
2
[Y ]. Hence, for 0
2

1
1, X
1
Y
X

2
Y proved.
Proposition II.4 The ffd X
1
Y classical fd X Y .
Proposition II.5 The ffd X
0
Y X does not functionally
determine Y .
III. STUDY OF FUZZY ARMSTRONGS
AXIOMS(F-ARMSTRONGS AXIOMS)
It is well-known that Armstrongs axioms form an important
basis of classical relational database. These axioms have
been extended by Al-Hamouz and Biswas [1] for a fuzzy
environment and named as Fuzzy Armstrongs Axioms (f-
Armstrongs Axioms). The f-Reexive rule, f-Augmentation
rule and f-Transitive rule proposed in ref. [1] are as follows:
Proposition III.1 (f-Armstrongs Axioms):
1) f-Reexive rule: If Y X, then X
()

Y ;
2) f-Augmentation rule: If X
()

Y , then XZ
()

Y Z;
3) f-Transitive rule: If X
()

Y and Y
()

Z, then
X
()

Z.(Proof: See ref. [1])


In the present work the f-Transitive rule of [1] has been
modied as follows:
Proposition III.2 Modied f-Transitive rule: If X

1
Y and
Y
2
Z, then X
min(
1
,
2
)
Z.
Proof: Assume that both the ffds X
1
Y and Y
2
Z
holds in the relation r(R).
Case I: When
1

2
then min(
1
,
2
) =
2
.
Given that X
1
Y and 0
2

1
1. So, using
Proposition II.3 we get
X

2
Y (i)
MISHRA & GHOSH, A STUDY OF FUZZY RELATIONAL DATABASE 47
From (i) we can write
t
1
[X]

2
t
2
[X] t
1
[Y ]

2
t
2
[Y ] (ii)
Again since ffd Y
2
Z is true, so we have
t
1
[Y ]

2
t
2
[Y ] t
1
[Z]

2
t
2
[Z] (iii)
Combining (ii) and (iii), we get
t
1
[X]

2
t
2
[X] t
1
[Z]

2
t
2
[Z]
which implies X

2
Z.
Hence for
2
= min(
1
,
2
), if X

1
Y and Y

2
Z
then X
2
Z proved.
Case II: When
2

1
, then min(
1
,
2
) =
1
.
Given that Y

2
Z and 0
1

2
1. So, using
Proposition II.3 we get
Y

1
Z (i)
From (i) we can write
t
1
[Y ]
1
t
2
[Y ] t
1
[Z]
1
t
2
[Z] (ii)
Again since ffd X

1
Y is true, so we have
t
1
[X]
1
t
2
[X] t
1
[Y ]
1
t
2
[Y ] (iii)
Combining (iii) and (ii), we get
t
1
[X]

1
t
2
[X] t
1
[Z]

1
t
2
[Z]
which implies X
1
Z.
Hence for
1
= min(
1
,
2
), if X
1
Y and Y
2
Z,
then X

1
Z proved.
Combining the above two cases, we can say if X
1
Y
and Y

2
Z, then X
min(
1
,
2
)
Z. Hence proved.
Consequently f-Union, f-Pseudotransitive, and f-
Decomposition rules of [1] have also been modied as
follows:
Proposition III.3 Modied f-Union rule: If X
1
Y and
X

2
Z, then X
min(1,2)
Y Z.
Proof: Given that
X

1
Y (i)
X

2
Z (ii)
From (i) we can write
X
1
XY (iii)[using f-Augmentation rule].
From (ii) we can write
XY

2
Y Z(iv)[using f-Augmentation rule].
From (iii) and (iv) using modied f-Transitive rule, we get
X
min(1,2)
Y Z. Hence proved.
Proposition III.4 Modied f-Pseudotransitive rule: If X

1
Y and WY
2
Z, then WX
min(
1
,
2
)
Z.
Proof: Given that
X

1
Y, (i)
WY

2
Z. (ii)
From (i) we can write
WX

1
WY. (iii)[using f-Augmentation rule]
From (iii) and (ii) using modied f-Transitive rule, we get
WX
min(
1
,
2
)
Z. Hence proved.
Proposition III.5 Modied f-Decomposition rule: If X

Y Z, then X

Y and X

Z.
Proof: Given that
X

Y Z (i)
Now Y Y Z. By f-Reexive rule, we have
Y Z

Y (ii)
From (i) and (ii) using modied f-Transitive rule, we get
X
min(,)
Y , i.e., X

Y [ min(, ) = ].
Similarly for given X

Y Z and Z Y Z, we get
X

Z. Hence proved.
In the following sections, we have referred our modied
rules as simple f-Transitive, f-Union, f-Pseudotransitive and
f-Decomposition rules.
IV. PARTIAL FUZZY FUNCTIONAL DEPENDENCY(PARTIAL
FFD)
After validation of Armstrongs Axioms in the fuzzy envi-
ronment with our present notion of ffd, let us dene partial
fuzzy functional dependency(partial ffd) as follows:
Denition IV.1 Partial fuzzy functional dependency: Y is
called partially fuzzy functionally dependent on X at -level
of choice, i.e., X

Y partially, iff X

Y and there exists


a non empty set X

X, such that, X

Y . The concept
of partial ffd then expresses the fact that after removal of an
attribute A from X, the dependency still holds, i.e., for an
attribute A X, X {A} still fuzzy functionally determines
Y at -level of choice.
Example IV.1: Let the relational schema be
R = {A, B, C, D, E} and the set of ffds F on R is
given by F = {ABC
0.75
Dand AC
0.8
D}. We prove
that the ffd ABC
0.75
D is a partial ffd.
Proof: Here ffd ABC
0.75
D holds which implies D
fuzzy functionally depends on ABC at 0.75-level of choice.
Again from the ffd AC
0.8
D, we say, D fuzzy functionally
depends on AC at 0.8-level of choice.
48 INTERNATIONAL JOURNAL OF COMPUTATIONAL COGNITION (HTTP://WWW.IJCC.US), VOL. 6, NO. 4, DECEMBER 2008
Now from the Proposition II.3, we have, AC
0.8
D
AC
0.75
D[ 0.75 < 0.8]. So, we comment that D is fuzzy
functionally dependent on AC at 0.75-level of choice.
We have the ffd ABC
0.75
D and we have proved the
existence of the ffd AC
0.75
D where AC ABC. Hence
ABC
0.75
D is a partial ffd. Hence proved.
V. FUZZY KEY
It is well known that in classical relational database key is
a special case of functional dependency. Let us now extend
the idea of classical key in the fuzzy environment to dene
fuzzy key with -level of choice where [0, 1] is a
choice parameter dened by the database designer. A formal
denition of fuzzy key is as follows:
Denition V.1 Fuzzy Key: Let K R and F be a set of ffds
for R. Then, K is called a fuzzy key of R at -level of choice
where [0, 1] iff K

R F and K

R is not a partial
ffd.
Example V.1: Let us assume a relation
schema R = (A, B, C, D) and a set of ffds
F = {A
0.75
B, A
0.8
C, A
0.7
D} of R. Find a fuzzy
key of R.
Solution: Given
A
0.75
B, (i)
A
0.8
C, (ii)
A
0.7
D. (iii)
Applying f-Union rule on (i) and (ii), we get
A
0.75
BC. (iv)
Again applying f-Union rule on (iii) and (iv), we get
A
0.7
BCD. (v)
Also, we know
A
1
A. (vi)[using Proposition II.4]
From (v) and (vi) using f-Union rule, we get A
0.7
ABCD,
i.e., A
0.7
R. A is a fuzzy key of R at 0.7-level of choice.
VI. FUZZY CLOSURE OF AN ATTRIBUTE OR A SET OF
ATTRIBUTES
Given a set of ffds for a relation, a fuzzy key of that relation
can be found utilizing the concept of fuzzy closure of an
attribute or a set of attributes. Fuzzy closure of an attribute
or a set of attributes X denoted by X
+
is the set of attributes
which are fuzzy functionally determined by the attribute(s) X.
Below we explain an algorithm to nd the fuzzy closure of an
attribute or a set of attributes.
Algorithm VI.1: Computation of fuzzy closure of an attribute
or a set of attributes.
Input: A fuzzy relational database R, a set of ffds F on R
and an attribute or a set of attributes X.
Output: A set of attributes X
+
, the fuzzy closure set of X.
Method:
Let X
+
= X; X
+
holds all attributes of X with fuzzy
membership value 1 and the attributes that do not belong to
X with membership value 0.
repeat
old X
+
= X
+
;
for each ffd Y

Z in F
do
if (Y, ) X
+
then for each attribute Z
i
of Z take fuzzy
union of X
+
and (Z
i
, ) and store the result in X
+
, i.e., if
(Y, ) X
+
then X
+
= X
+

fuzzy
(Z
i
, ) [(Y, ) X
+
holds if is less than or equal the membership value already
stored in X
+
of Y ].
end for
until (X
+
= old X
+
)
Example VI.1: Let R(A, B, C, D) be
a relational schema and a set of ffds
F = {A
0.75
B, A
0.8
B, A
0.6
B, B
0.6
C, A
0.85
CD}
on R. Find fuzzy closure of A.
Here initially A
+
= {(A, 1), (B, 0), (C, 0), (D, 0)}; In
repeat until loop of Algorithm VI.1,
Step 1 of repeat loop :
old A
+
= {(A, 1), (B, 0), (C, 0), (D, 0)};
In for loop we have the following execution :
Step1:
consider ffd A
0.75
B, here (A, 0.75) A
+
[ 0.75 < 1]
so compute A
+
= A
+

fuzzy
(B, 0.75), which
A
+
= {(A, 1), (B, 0.75), (C, 0), (D, 0)}.
Step2:
for ffd A
0.8
B since (A, 0.8) A
+
,
compute A
+
= A
+

fuzzy
(B, 0.8), i.e.,
A
+
= {(A, 1), (B, 0.8), (C, 0), (D, 0)}.
Step3:
for ffd A
0.6
B since (A, 0.6) A
+
,
compute A
+
= A
+

fuzzy
(B, 0.6), i.e.,
A
+
= {(A, 1), (B, 0.8), (C, 0), (D, 0)} no change since
fuzzy union contain the maximum membership value.
Step4:
for ffd B
0.6
C since (B, 0.6) A
+
,
compute A
+
= A
+

fuzzy
(C, 0.6), i.e.,
A
+
= {(A, 1), (B, 0.8), (C, 0.6), (D, 0)}.
Step5:
for ffd A
0.85
CD since (A, 0.85) A
+
,
compute A
+
= A
+

fuzzy
{(C, 0.85), (D, 0.85)}, i.e.,
A
+
= {(A, 1), (B, 0.8), (C, 0.85), (D, 0.85)}.
end of for loop
Here, old A
+
= A
+
.
Go to repeat loop
Step 2 of repeat loop:
Now old A
+
= {(A, 1), (B, 0.8), (C, 0.85), (D, 0.85)}.
In for loop again considering all the ffds we have no change
in A
+
.
MISHRA & GHOSH, A STUDY OF FUZZY RELATIONAL DATABASE 49
end of for loop
Here old A
+
= A
+
.
So, we exit from repeat loop.
Algorithm terminated.
Finally we got fuzzy closure of A, i.e.,
A
+
= {(A, 1), (B, 0.8), (C, 0.85), (D, 0.85)}.
VII. REAL LIFE EXAMPLES
Below a real life example is given to discuss the above
theory.
Example VII.1: Consider a FACULTY relation schema as
FACULTY(Name, Address, Age, Experience, Salary) and
F ={Name Address, Name Age,
Age
0.8
Experience, Experience
0.7
Salary}
is a set of fds and ffds on FACULTY.
1) To nd a fuzzy key of the relation FACULTY.
2) To nd fuzzy closure of the attribute Age.
Solution:
(i) Given that
Name Address Name
1
Address (i)
[Using Proposition II.5]
Name Age Name
1
Age (ii)
[Using Proposition II.5]
Age
0.8
Experience (iii)
Experience
0.7
Salary (iv)
From (ii) and (iii) using f-Transitive rule, we get
Name
0.8
Experience (v)
From (v) and (iv) using f-Transitive rule, we get
Name
0.7
Salary (vi)
Now applying f-Union rule on (i), (ii), (v), (vi), we get
Name
0.7
Address Age Experience Salary (vii)
and we know
Name
1
Name (viii)
Form (vii) and (viii) using f-Union rule, we get
Name
0.7
Name Address Age Experience Salary,
i.e, Name
0.7
FACULTY .
Hence, Name is a fuzzy key of FACULTY relation at the
0.7 level of choice.
(ii) We apply Algorithm VI.1 to nd the fuzzy closure of Age.
Initially,
(Age)
+
={(Age, 1), (name, 0), (Address, 0),
(Experience, 0), (Salary, 0)}
Step 1:
Old(Age)
+
={(Age, 1), (name, 0), (Address, 0),
(Experience, 0), (Salary, 0)}.
1) For the ffd Name
1
Address, (Age)
+
will have no
change [ (Name, 1) (Age)
+
].
2) Similarly for the ffd Name
1
Age, (Age)
+
will remain
same.
3) For the ffd Age
0.8
Experience, compute
(Age)
+
= (Age)
+

fuzzy
(Experience, 0.8)
[ (Age, 0.8) (Age)
+
], i.e.,
(Age)
+
={(Age, 1), (name, 0), (Address, 0),
(Experience, 0.8), (Salary, 0)}.
4) For the ffd Experience
0.7
Salary, com-
pute (Age)
+
= (Age)
+

fuzzy
(Salary, 0.7)
[ (Experience, 0.7) (Age)
+
], i.e.,
(Age)
+
={(Age, 1), (name, 0), (Address, 0),
(Experience, 0.8), (Salary, 0.7)}.]
Here, old (Age)
+
= (Age)
+
.
Step 2:
Now
old(Age)
+
= {(Age, 1), (name, 0), (Address, 0),
(Experience, 0.8), (Salary, 0.7)}
Next, again considering all the ffds we have no change in
(Age)
+
. Here old (Age)
+
= (Age)
+
. Hence nally,
(Age)
+
={(Age, 1), (name, 0), (Address, 0),
(Experience, 0.8), (Salary, 0.7)},
i.e.,
(Age)
+
= {(Age, 1), (Experience, 0.8), (Salary, 0.7)}.
VIII. REDUNDANT FUZZY FUNCTIONAL
DEPENDENCY(REDUNDANT FFD)
Let F be a set of ffds of a relation R. Now an ffd is said to
be redundant if that ffd can be deduced from the remaining
ffds of F.
Example VIII.1 Suppose R(A, B, C) be a relational schema
and a set of ffds F = {A
0.7
B, B
0.8
C, A
0.6
C} on R.
Here ffd A
0.6
C is redundant. Because ffd A
0.6
C can be
deduced from ffds A
0.7
B and B
0.8
C.
Proof: By the f-Transitive rule of ffds,
A
0.7
B, B
0.8
C A
0.7
C (i)
Again, using Proposition II.3,
A
0.7
C A
0.6
C (ii)
From (i) and (ii), we get
A
0.7
B, B
0.8
C A
0.6
C.
Hence, the ffd A
0.6
C is redundant.
50 INTERNATIONAL JOURNAL OF COMPUTATIONAL COGNITION (HTTP://WWW.IJCC.US), VOL. 6, NO. 4, DECEMBER 2008
IX. CONCLUSIONS
It is widely recognized that many attributes in real life
applications are fuzzy in nature and thus study of fuzzy
functional dependency in fuzzy relational databases play a
vital rule. In the present work, the new notion of -ffd
introduced in [1], which is based on an equivalence relation,
has been studied in details with certain modications in f-
Armstrongs axioms. I have also introduced the ideas of partial
ffd, fuzzy key, computation of fuzzy closure of attribute(s) and
redundant ffd. All these ideas have been validated with a real
life example.
As future work, I want to extend this paper to study
the Normalization, Lossless join decomposition, Dependency
Preservation and Multivalued Dependency using -ffd as
dened in [1] which constitute an important part of a good
fuzzy relational database design.
REFERENCES
[1] S. Al-Hamouz and R. Biswas. Fuzzy functional dependencies in relational
databases. International Journal of Computational Cognition, 4(1):3943,
2006.
[2] B. Bhunia and P. Niyogi. Lossless join property in fuzzy relational
databases. Data of Knowledge Engineering, 11:109124, 1993.
[3] P. B. Buckles and F. E. Petry. A fuzzy representation of data for relational
databases. Fuzzy Sets and Systems, 7:213226, 1982.
[4] G. Chen, E. E. Kerre, and J. Vandenbulcke. Normalization based on ffd
in a fuzzy relational data model. Information Systems, 21(3):299310,
1996.
[5] E. Codd. A relational model for large shared data banks. Comm. of ACM,
13:377387, 1970.
[6] F. E. Petry. Fuzzy Databases: Principles and Applications. Kluwer
Academic Pub., Boston, 1996.
[7] K. V. S. V. N. Raju and A. K. Majumdar. Fuzzy functional dependencies
and lossless join decomposition of fuzzy relational database systems.
ACM Transactions on Database Systems, 13(2):129166, 1988.
[8] S. Shenoi, A. Melton, and L.T. Fan. Functional dependencies and normal
forms in fuzzy relational database model. Information Sciences, 60:128,
1992.

You might also like