Information Fusion xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Information Fusion
journal homepage: www.elsevier.com/locate/inffus

Construction of admissible linear orders for interval-valued Atanassov

intuitionistic fuzzy sets with an application to decision making
L. De Miguel a, H. Bustince a,c,⇑, J. Fernandez a, E. Induráin b, A. Kolesárová d, R. Mesiar d,e,f
a
Departamento de Automática y Computación, Universidad Pública de Navarra, Campus Arrosadia s/n, 31006 Pamplona, Spain
b
Departamento de Matemáticas, Universidad Pública de Navarra, Campus Arrosadia s/n, 31006 Pamplona, Spain
c
Institute of Smart Cities, Universidad Publica de Navarra, Campus Arrosadia s/n, 31006 Pamplona, Spain
d
Institute of Information Engineering, Automation and Mathematics, Slovak University of Technology, 81237 Bratislava, Slovakia
e
Slovak University of Technology, Radlinskeho 11, Bratislava, Slovakia
f
Institute of Information Theory and Automation, Academy of Sciences of the Czech Republic, 18208 Prague, Czech Republic

a r t i c l e i n f o a b s t r a c t

Article history: In this work we introduce a method for constructing linear orders between pairs of intervals by using
Received 10 November 2014 aggregation functions. We adapt this method to the case of interval-valued Atanassov intuitionistic fuzzy
Received in revised form 23 March 2015 sets and we apply these sets and the considered orders to a decision making problem.
Accepted 28 March 2015
Ó 2015 Elsevier B.V. All rights reserved.
Available online xxxx

Keywords:
Interval-valued Atanassov intuitionistic
fuzzy set
Interval linear order
Interval-valued Atanassov intuitionistic
multi-expert decision making

1. Introduction 1. If we want to model environments where there exist non-

comparable elements, it will be enough to use classical partial
In decision making problems it may happen that, after the orders between intervals. This is not the case in this work.
exploitation phase, the best alternatives are equally ranked and it 2. If we must represent ignorance [8] associated to the datum
is not possible to decide which one is the best. It has been noticed given by an expert, we can understand the length of the inter-
[1] that these troubles often appear when the entries of the consid- vals as a representation of such ignorance. If, in these cases,
ered fuzzy preference matrix are close to 0.5, that is, when the we need to be able to compare any two data, then we can use
experts have doubts about their preferences of some alternatives any of the linear orders we consider here.
over the others. In this situation, the systematic use of extensions
of fuzzy sets has been shown to be a really useful tool [2]. Among Once the decision of using IVAIFSs to deal with a decision mak-
those fuzzy sets, interval-valued fuzzy sets (IVFSs) [3–5] or, ing problem has been reached, we should choose, accordingly, a
equivalently, Atanassov intuitionistic fuzzy sets (AIFs) [6] play linear order between pairs of intervals. In this way, we will select
indeed a crucial role. as the best option the alternative which is associated to the largest
In some special cases, despite the fact of using IVFSs and AIFs, pair of intervals, with respect to the considered linear order.
still remain problems that are similar to those encountered in Moreover, in decision making problems we must also aggregate
the previous ones. For these new last situations we may use the the information furnished by the experts by means of aggregation
interval-valued Atanassov intuitionistic fuzzy sets (IVAIFSs) [7]. functions [9–11].
Besides, the use of intervals to represent membership and non- All these considerations have led us to aim the following
membership has, from our point of view, a double advantage: objectives:

⇑ Corresponding author at: Departamento de Automática y Computación, (1) To use aggregation functions for building linear orders for
Universidad Pública de Navarra, Campus Arrosadia s/n, 31006 Pamplona, Spain. pairs of intervals whose end-points belong to the unit
E-mail addresses: [email protected] (L. De Miguel), bustince@ interval.
unavarra.es (H. Bustince), [email protected] (J. Fernandez), steiner@ (2) To study methods for constructing linear orders on the set of
unavarra.es (E. Induráin), [email protected] (A. Kolesárová), mesiar@math.
sk (R. Mesiar).
IVAIFSs.

http://dx.doi.org/10.1016/j.inffus.2015.03.004
1566-2535/Ó 2015 Elsevier B.V. All rights reserved.

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
2 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx

(3) To deal with the exploitation phase of decision making prob- Definition 2.1 [18]. An order  on Lð½0; 1Þ is said to be admissible
lems through IVAIFSs, by using the previously built linear if it is linear and refines the order 2 , i.e., it is a linear order
orders. satisfying that for all x; y 2 Lð½0; 1Þ such that x 2 y it holds x  y.

The structure of this paper is the following. In Section 2 we

Example 2.1. The lexicographic orders on Lð½0; 1Þ, given by
introduce the notation and recall some well-known notions. In
Sections 3,4, we construct two classes of linear orders between
 x lex1 y if and only if ðx < yÞ or ðx ¼ y and x 6 yÞ (lexicographic-1
pairs of intervals. Section 5 contains an application of our theoreti-
cal results to group decision making. In particular, we provide two order), and
algorithms. Some concluding remarks as well as suggestions for  x lex2 y if and only if ðx < yÞ or ðx ¼ y and x 6 yÞ (lexicographic-2
further research close the paper. order), are admissible.

2. Previous concepts and results 2.2. Extensions of fuzzy sets

We start by recalling some well-known concepts that will be Definition 2.2 [6]. Let U be a nonempty set usually called a
useful for subsequent developments throughout the paper. universe. An Atanassov’s Intuitionistic Fuzzy Set (AIFS) F over U is
given by
2.1. On orders and partially ordered sets
F ¼ fhu; lF ðuÞ; mF ðuÞiju 2 Ug
Given a partially ordered set (poset) ðP; Þ, we say that
where lF : U ! ½0; 1 defines the membership degree of the element
(a) 1P is the top of the poset if for all x 2 P it holds x  1P . u 2 U to F and mF : U ! ½0; 1 defines its nonmembership degree to
(b) 0P is the bottom of the poset if for all x 2 P it holds 0P  x. the same set F. Besides, the functions lF and mF satisfy that, for all
u 2 U; lF ðuÞ þ mF ðuÞ 6 1.
In case they exist, 1P and 0P are unique. The pair ðlF ðuÞ; mF ðuÞÞ is called an intuitionistic pair, Lð½0; 1Þ
Let Kð½0; 1Þ  R2 be given by being the set of all possible intuitionistic pairs, i.e.,
Kð½0; 1Þ ¼ fðx; xÞ 2 ½0; 1  ½0; 1j x 6 xg; Lð½0; 1Þ ¼ faja ¼ ða1 ; a2 Þ; a1 ; a2 2 ½0; 1 and a1 þ a2 6 1g:
and let Lð½0; 1Þ be the set of all closed subintervals of the unit inter- In [6], Atanassov introduced a partial order in the universe of
val, that is AIFSs.
Lð½0; 1Þ ¼ fxjx ¼ ½x; x such that 0 6 x 6 x 6 1g:
Definition 2.3. Let F 1 ; F 2 be two AIFSs. According to the order
There is a straightforward bijection i : Kð½0; 1Þ ! Lð½0; 1Þ given given by Atanassov in [6]
by iððx; xÞÞ ¼ ½x; x ¼ x. Through this bijection, the partial order on
R2 ; ða; bÞ2 ðc; dÞ if and only if a 6 c and b 6 d induces an equiva- F 1 6 F 2 if and only if for all u 2 U; lF 1 ðuÞ 6 lF 2 ðuÞ and mF 1 ðuÞ
lent partial order on Lð½0; 1Þ, namely, P mF 2 ðuÞ:
x 2 y iff x 6 y and x 6 y: ð1Þ

In this way, ðLð½0; 1Þ; 2 Þ is a poset whose bottom and top are, Definition 2.4 [7]. Let U be a universe. An Interval-Valued
respectively, 0 ¼ ½0; 0 and 1 ¼ ½1; 1. In fact, the bijection above is Atanassov Intuitionistic Fuzzy Set (IVAIFS) G over U is given by
a lattice isomorphism.1
We refer as ðLð½0; 1ÞÞ2 , to the universe of pairs of intervals, that G ¼ fhu; mG ðuÞ; nG ðuÞiju 2 Ug
is,
where mG : U ! Lð½0; 1Þ defines the membership degree of the ele-
ðLð½0; 1ÞÞ2 ¼ fðx; yÞ ¼ ð½x; x; ½y; yÞ with x; x; y; y 2 ½0; 1g: ment u 2 U to F and nG : U ! Lð½0; 1Þ defines its nonmembership
degree to the same universe U. Moreover, for all u 2 U, the sum of
Similarly to what happens in the case of R2 and Lð½0; 1Þ, the par-
the upper boundary values of mG ðuÞ and nG ðuÞ must be lower than
tial order on R4 , given by ða1 ; b1 ; c1 ; d1 Þ4 ða2 ; b2 ; c2 ; d2 Þ if and only or equal to 1.
if a1 6 a2 and b1 6 b2 and c1 6 c2 and d1 6 d2 , also induces an
equivalent partial order 4 on ðLð½0; 1ÞÞ2 , given by The pair ðmG ðuÞ; nG ðuÞÞ is called an interval-valued intuitionistic
pair, being LIV ð½0; 1Þ the set of all possible interval-valued intuitio-
ðx1 ; y1 Þ 4 ðx2 ; y2 Þ if and only if x1 6 x2 and x1 6 x2 and y1 nistic pairs, i.e.,
6 y2 and y1 6 y2 : ð2Þ LIV ð½0; 1Þ ¼ fðx; yÞ; with x; y 2 Lð½0; 1Þ and x þ y 6 1g:
2
In this way, ððLð½0; 1ÞÞ ; 4 Þ becomes a poset whose bottom and top
are, respectively, ð0; 0Þ ¼ ð½0; 0; ½0; 0Þ and ð1; 1Þ ¼ ð½1; 1; ½1; 1Þ.
Remark 1. Note that LIV ð½0; 1Þ consists of special types of inter-
1
This kind of sets, namely Kð½0; 1Þ and Lð½0; 1Þ have already been used, suitably
vals, while ðLð½0; 1ÞÞ2 is a set of all possible intuitionistic pairs.
equipped with some order and latticial structure [12,13], to construct some universal
codomain where it was possible to represent different kinds of orderings as, e.g., total
preorders, interval-orders and semiorders by means of a single function that
Definition 2.5. Let G1 ; G2 be two IVAIFSs. According to the order
preserves the ordinal structure. The bijection i : Kð½0; 1Þ ! Lð½0; 1Þ has also been
considered in those approaches, and some other similar bijections and/or latticial given by Atanassov in [7], G1  G2 if and only if, for all u 2 U,
isomorphism as well as order isotonies have also been introduced accordingly. By the
way, another universal codomain to represent different kinds of orderings, which is mG1 ðuÞ 2 mG2 ðuÞ and nG2 ðuÞ 2 nG1 ðuÞ;
essentially equivalent to Kð½0; 1Þ, consists of triangular and symmetric fuzzy
numbers. For further information see [14–17]. where 2 is the partial order on Lð½0; 1Þ given in Eq. (1).

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx 3

2.3. Aggregation functions  The wider the first interval, the wider the rectangle.
 The wider the second interval, the higher the rectangle.

Definition 2.6. Given a poset ðP; P Þ with bottom 0P and top 1P , an As a consequence, the area of the rectangle will be directly pro-
aggregation function M on P w.r.t the order P (also known as an portional to the width of the intervals. Furthermore, for any
P -aggregation function) is a mapping M : Pn ! P satisfying z1 ; z2 2 ðLð½0; 1ÞÞ2 ; z1 4 z2 if and only if each corner of the rectan-
gle of z2 is located above and on the right side of its corresponding
 Mð0P ; . . . ; 0P Þ ¼ 0P ; Mð1P ; . . . ; 1P Þ ¼ 1P , and corner in the rectangle z1 .
 Mðx1 ; . . . ; xn Þ P Mðy1 ; . . . ; yn Þ for ðx1 ; . . . ; xn Þ P ðy1 ; . . . ; yn Þ
Example 3.1. Let z1 ¼ ð½0:3; 0:6; ½0:2; 0:7Þ; z2 ¼ ð½0:5; 0:8; ½0:55;
where ðx1 ; . . . ; xn Þ P ðy1 ; . . . ; yn Þ holds if and only if xi P yi for all 0:9Þ; z3 ¼ ð½0:4; 0:5; ½0:3; 0:35Þ; z4 ¼ ð½0:1; 0:4; ½0:4; 0:6Þ. The
i 2 f1; . . . ; ng. intervals can be represented in the unit square ½0; 12 as in Fig. 1.
In that figure some visual interpretations can be drawn. For
This definition extends the usual one for the unit interval ½0; 1.
example, we have that the intervals of z1 are wider than those of
For further information see [19].
any other zi , since its area is significantly greater. Alternatively, we
have that zi 4 z2 for i 2 f1; 3; 4g, since the corners of z2 are located
Proposition 2.1 [18]. Let B1 ; B2 : ½0; 12 ! ½0; 1 be two continuous
above and on the right side w.r.t the other rectangles. Similarly, we
aggregation functions, such that for all ðp1 ; p2 Þ; ðq1 ; q2 Þ 2 Kð½0; 1Þ, the
can deduce that z1 ; z3 and z4 are incomparable in terms of 4 .
equalities B1 ðp1 ; p2 Þ ¼ B1 ðq1 ; q2 Þ and B2 ðp1 ; p2 Þ ¼ B2 ðq1 ; q2 Þ only hold
provided that ðp1 ; p2 Þ ¼ ðq1 ; q2 Þ. In [18], Bustince et al. introduced a construction method of admis-
The order B1 ;B2 on Lð½0; 1Þ, given by sible orders on Lð½0; 1Þ by using two aggregation functions. Such
method can also be generalized to handle elements in ðLð½0; 1ÞÞ2 .
x B1 ;B2 y if and only if B1 ðx; xÞ < B1 ðy; yÞ or else ðB1 ðx; xÞ
¼ B1 ðy; yÞ and B2 ðx; xÞ 6 B2 ðy; yÞÞ; Proposition 3.1. Let A ¼< A1 ; A2 ; A3 ; A4 > be four aggregation func-
tions,2 Ai : ½0; 14 ! ½0; 1 such that for all ðp; qÞ; ðr; sÞ 2 ðLð½0; 1ÞÞ2 the
is an admissible order on Lð½0; 1Þ. equalities Ai ðp; p; q; qÞ ¼ Ai ðr; r; s; sÞ for all i ¼ f1; . . . 4g only hold if
ðp; qÞ ¼ ðr; sÞ.
The following results can be found in [9,11,20,21].
An admissible order A on ðLð½0; 1ÞÞ2 can be defined as follows
ðx1 ; y1 Þ A ðx2 ; y2 Þ if and only if one of the (mutually exclusive)
Definition 2.7. A function T : ½0; 12 ! ½0; 1 is called a t-norm if it
following conditions is satisfied.
is symmetric, associative, increasing with respect to the
order 6 and Tðx; 1Þ ¼ x for all x 2 ½0; 1.
(i) A1 ðx1 ; x1 ; y1 ; y1 Þ < A1 ðx2 ; x2 ; y2 ; y2 Þ;
(ii) A1 ðx1 ; x1 ; y1 ; y1 Þ ¼ A1 ðx2 ; x2 ; y2 ; y2 Þ and A2 ðx1 ; x1 ; y1 ; y1 Þ <
Definition 2.8. A function S : ½0; 12 ! ½0; 1 is called a t-conorm if A2 ðx2 ; x2 ; y2 ; y2 Þ;
it is symmetric, associative, increasing with respect to the (iii) A1 ðx1 ; x1 ; y1 ; y1 Þ ¼ A1 ðx2 ; x2 ; y2 ; y2 Þ and A2 ðx1 ; x1 ; y1 ; y1 Þ ¼
order 6 and Sðx; 0Þ ¼ x for all x 2 ½0; 1. A2 ðx2 ; x2 ; y2 ; y2 Þ and A3 ðx1 ; x1 ; y1 ; y1 Þ < A3 ðx2 ; x2 ; y2 ; y2 Þ;
A strictly decreasing and continuous function n : ½0; 1 ! ½0; 1 (iv) A1 ðx1 ; x1 ; y1 ; y1 Þ ¼ A1 ðx2 ; x2 ; y2 ; y2 Þ and A2 ðx1 ; x1 ; y1 ; y1 Þ ¼
such that nð0Þ ¼ 1 and nð1Þ ¼ 0 is called a strict negation. If, in A2 ðx2 ; x2 ; y2 ; y2 Þ and A3 ðx1 ; x1 ; y1 ; y1 Þ ¼ A3 ðx2 ; x2 ; y2 ; y2 Þ and
addition, it is involutive (that is, nðnðxÞÞ ¼ x for all x 2 ½0; 1), then A4 ðx1 ; x1 ; y1 ; y1 Þ 6 A4 ðx2 ; x2 ; y2 ; y2 Þ.
n is said to be a strong negation. A t-norm T is dual to a t-conorm
S (and vice versa) with respect to a strong negation n if
Proof. The order A refines 4 since every Ai is an aggregation
Tðx; yÞ ¼ nðSðnðxÞ; nðyÞÞÞ for all x; y 2 ½0; 1.
function. Moreover, the linearity is assured since the four equalities
of Ai only hold simultaneously if ðx1 ; y1 Þ ¼ ðx2 ; y2 Þ. The transitivity
3. Admissible orders on ðLð½0; 1ÞÞ2 follows from the transitivity of the standard order on ½0; 1. h

Although a partial order is enough to define aggregation func-

tions, some special classes of aggregations actually require to have Remark 2. Notice that any permutation of the aggregation func-
at hand a linear order. Examples of such classes are Choquet inte- tions Ai also produces an admissible order different from the for-
grals and Sugeno integrals. The order given by Atanassov for mer one.
IVAIFSs is partial, which is a undeniable handicap in the adaptation
of such classes of aggregation operators to the IVAI setup. In this Remark 3. In [18] it was proven that an admissible order on
section we define the admissible linear orders on ðLð½0; 1ÞÞ2 , gener- Kð½0; 1Þ cannot be induced by a single function. Clearly, this result
alizing the concept of admissible orders on Lð½0; 1Þ. also holds true since we are working in a larger space.
Henceforward, we use the order generated by four aggregation
Definition 3.1. An order  on ðLð½0; 1ÞÞ2 is said to be admissible if
functions (in Proposition 3.1). Thus, all the ideas to be introduced
it is a linear and refines the order 4 in Eq. (2), i.e., it is
till the end of this section refer to such family of admissible orders
linear order satisfying that for all ðx1 ; y1 Þ; ðx2 ; y2 Þ 2 ðLð½0; 1ÞÞ2 ; named 4-admissible.
ðx1 ; y1 Þ 4 ðx2 ; y2 Þ implies ðx1 ; y1 Þ  ðx2 ; y2 Þ.
Example 3.2. The lexicographic orders can be constructed from
The elements zi ¼ ðxi ; yi Þ 2 ðLð½0; 1ÞÞ2 can be visualized in a the four projections.
straightforward manner. Since xi ; yi 2 Lð½0; 1Þ, each pair of inter-
vals can be drawn as a rectangle for which the first interval lies
in the horizontal axis and the second interval lies in the vertical
one. In such a representation, the following statements hold true: 2
Warning: notice that here the order of appearance of the A0i s counts. See also Remark 2.

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
4 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx

Fig. 1. Pairs of intervals.

1. The standard lexicographic order: let Ai be the aggregation Proof. The functions Ai are weighted arithmetic means. Let
function that maps to the ith component (i.e. the ith projection). ð½x1 ; x1 ; ½y1 ; y1 Þ; ð½x2 ; x2 ; ½y2 ; y2 Þ 2 ðLð½0; 1ÞÞ2 , such that
In that case, ðx1 ; y1 Þ A ðx2 ; y2 Þ if and only if
 (x1 < x2 ), or ai x1 þ bi x1 þ ci y1 þ di y1 ¼ ai x2 þ bi x2 þ ci y2 þ di y2 ;
 (x1 ¼ x2 and x1 < x2 ), or
for i 2 f1; . . . ; 4g. Because of the regularity of D, both linear systems
 (x1 ¼ x2 ; x1 ¼ x2 and y1 < y2 ), or
have a unique and common solution, i.e., ðx1 ; x1 ; y1 ; y1 Þ ¼
 (x1 ¼ x2 ; x1 ¼ x2 ; y1 ¼ y2 and y1 6 y2 ).
ðx2 ; x2 ; y2 ; y2 Þ. The result now follows from Proposition 3.1. h

2. The reversed lexicographic order: let Ai be the aggregation func- Example 3.3. Let A contain the following aggregation functions:
tion that maps to the ð5  iÞth component (i.e. the ð5  iÞth pro-
jection). In that case, ðx1 ; y1 Þ A ðx2 ; y2 Þ if and only if  A1 ðx1 ; x1 ; y1 ; y1 Þ ¼ 38 x1 þ 38 x1 þ 18 y1 þ 18 y1 ;
 (y1 < y2 ), or  A2 ðx1 ; x1 ; y1 ; y1 Þ ¼ 10 5
x þ 20 3
x1 þ 20 2
y1 þ 20 y1 ;
20 1
 (y1 ¼ y2 and y1 < y2 ), or 1
 A3 ðx1 ; x1 ; y1 ; y1 Þ ¼ 20 x1 þ 10 8
x þ 20
20 1
1
y1 þ 20 y1 ;
 (y1 ¼ y2 ; y1 ¼ y2 and x1 < x2 ), or
 A4 ðx1 ; x1 ; y1 ; y1 Þ ¼ 14 x1 þ 14 x1 þ 14 y1 þ 14 y1 .
 (y1 ¼ y2 ; y1 ¼ y2 ; x1 ¼ x2 and x1 6 x2 ).
Since jDj ¼ 0:0069, the order generated by A, as in Proposition
3.1, is a 4-admissible order.
3. Any other permutation of the projections gives rise to an admis-
sible order where we compare the components in a predeter-
mined order.
Remark 4. Notice that the value of the determinant is close to 0
but this is due to the fact that all the elements of the matrix are
Proposition 3.2. Let A ¼< A1 ; A2 ; A3 ; A4 > be four aggregation func- smaller than 1.
tions given by The construction of admissible orders through a 4-tuple of
Ai ðx1 ; x1 ; y1 ; y1 Þ ¼ ai x1 þ bi x1 þ ci y1 þ di y1 ; weighted arithmetic means has an interesting geometrical inter-
pretation. If we consider A in the form of the corresponding four
with ai ; bi ; ci ; di 2 ½0; 1; ai þ bi þ ci þ di ¼ 1 and weighting vectors which generate A1 ; . . . ; A4 , i.e.,
 
 a1 b1 c1 d1  A R ¼ f< a1 ; b1 ; c1 ; d1 >; < a2 ; b2 ; c2 ; d2 >; < a3 ; b3 ; c3 ; d3 >;

a b2 c2 d2 
 2 < a4 ; b4 ; c4 ; d4 >g
jDj ¼   – 0:
 a3 b3 c3 d3 
 the condition in Proposition 3.2 means that R is a basis of the vector
a b4 c4 d4 
4
space R4 . Hence, to any basis R of R4 which consists of weighting
Then (and only then), the order generated by the aggregation functions vectors there is a unique admissible order A constructed by means
Ai is a 4-admissible order. of the corresponding weighted means.

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx 5

1 1  1 1 
Finally, after changing the basis, the values of interval-valued  2 2 0 0  2 2 0 0
   
intuitionistic pairs in the new basis, (which are now in ½0; 14 ), 0 1 0 0  1 1 0 0
  3 3 
are ordered through the standard lexicographic order.
jD1 j ¼   – 0; jD2 j ¼  –0
0 0 1 1   0 0 12 12 
 2 2   
0 0 0 1 0 0 1 2 
3 3
Proposition 3.3. Let a tuple A ¼ hA1 ; . . . ; A4 i of aggregation functions
generate an admissible order A . Let Bi : ½0; 12 ! ½0; 1; i 2 f1; . . . ; 4g generate the same order.
be four aggregation functions such that
4. IVAIF-admissible order on LIV ð½0; 1Þ
 Ai ðx; x; y; yÞ ¼ Bi ðx; xÞ for i 2 f1; 2g, and
 Aj ðx; x; y; yÞ ¼ Bj ðy; yÞ for j 2 f3; 4g. The admissible orders defined in Section 3 refine the partial
order 4 . However, any of them could also refine the partial order
Then, ðx1 ; y1 Þ A ðx2 ; y2 Þ if and only if given by Atanassov for IVAIFS [7]. In this section, we define a new
family of linear orders with a crucial additional feature, namely,
(i) (x1 B1 ;B2 x2 ), or they refine Atanassov’s partial order.
(ii) (x1 ¼ x2 and y1 B3 ;B4 y2 ), We remind the reader that in Atanassov’s partial order, given
two elements ðx1 ; y1 Þ; ðx2 ; y2 Þ 2 LIV ð½0; 1Þ,
where Bi ;Bj is the order on Lð½0; 1Þ generated in Proposition 2.1. ðx1 ; y1 Þ  ðx2 ; y2 Þ if and only if x1 6 x2 ; x1 6 x2 ; y1
P y2 ; and y1 P y2 : ð3Þ
Proof. It is straightforward.
Notice that, if we use B1 ¼ B3 and B2 ¼ B4 , the result is a Definition 4.1. An order  on LIV ð½0; 1Þ is said to be an IVAIF-
4-admissible order where we combine the standard lexicographic admissible order if it is a linear order and refines the partial order
order with the order B1 ;B2 . The resulting order acts as follows: first given by Atanassov for IVAIFS (Eq. (3)).
we compare the intervals using B1 ;B2 and, only if they are equal,
Notice that, if we have an IVAIF-admissible order on LIV ð½0; 1Þ,
we compare the second interval with that same order (B1 ;B2 ). For
as in Definition 4.1, then the bottom of ðLIV ð½0; 1Þ; Þ is ð0; 1Þ and
instance, the standard lexicographic order can be seen as the
the top is ð1; 0Þ.
composition of the lexicographic-1 order between intervals
As in Section 3, we can generate a visualization of the elements
combined with itself.
Alternatively, notice that, if Ai ðx; x; y; yÞ ¼ Bi ðy; yÞ for i 2 f1; 2g, in LIV ð½0; 1Þ  ðLð½0; 1ÞÞ2 that captures the behaviour of the admis-
sible orders in Definition 4.1. Following the visualization rules in
and Aj ðx; x; y; yÞ ¼ Bj ðx; xÞ for j 2 f3; 4g, then the resulting order is
Fig. 1 we have that, for any two elements z1 ; z2 in
also 4-admissible. LIV ð½0; 1Þ; z1  z2 if and only if the corners of z2 are individually
A well-known class of binary aggregation functions is that of located below and to the right of those of z1 . For example, in
Atanassov’s operators Ka given by Ka ða; bÞ ¼ a þ aðb  aÞ with Fig. 2, we have given z1 ¼ ð½0:1; 0:4; ½0:1; 0:6Þ; z2 ¼ ð½0:3; 0:55;
a 2 ½0; 1. ½0:05; 0:25Þ; z3 ¼ ð½0:05; 0:2; ½0:15; 0:25Þ 2 LIV ð½0; 1Þ. Visually, it is
In our particular case, the inputs being intervals, an Atanassov’s evident that z1  z2 and z3  z2 , but also that z1 and z3 are not
operator acting on the endpoints of the intervals yields a point comparable with the partial order in Definition 4.1. Notice that,
inside the corresponding intervals. h in this visualization, no element is allowed to be in the grey zone
of the rectangle in Fig. 2 due to the restrictions in the definitions
Example 3.4. Let a1 ; a2 ; a3 ; a4 2 ½0; 1, with a1 – a2 and a3 – a4 . Let of the membership and nonmembership degrees in an interval-val-
A ¼< A1 ; . . . ; A4 > be four aggregation functions given by ued intuitionistic pair.
In the sequel, two different constructions of IVAIF-admissible
orders are introduced.
 Ai ðx1 ; x1 ; y1 ; y1 Þ ¼ Kai ðx1 ; x1 Þ, for i 2 f1; 2g, and
 Aj ðx1 ; x1 ; y1 ; y1 Þ ¼ Kaj ðy1 ; y1 Þ, for j 2 f3; 4g.
Proposition 4.1. Let B ¼< B1 ; B2 ; B3 ; B4 > be four aggregation func-
tions Bi : ½0; 14 ! ½0; 1 which generate the orders B1 ;B2 and B3 ;B4 on
The tuple A generates a 4-admissible order that renders in
ðx1 ; y1 Þ A ðx2 ; y2 Þ if and only if Lð½0; 1Þ as in Proposition 2.1. Then the order relation IV B given by

ðx1 ; y1 Þ IV
B ðx2 ; y 2 Þ if and only if x1 B1 ;B2 x2 or ðx1
 (x1 Ka1 ;Ka2 x2 ), or
¼ x2 and y2 B3 ;B4 y1 Þ;
 (x1 ¼Ka1 ;Ka2 x2 and y1 Ka3 ;Ka4 y2 ).
is an IVAIF-admissible order.
From the construction in Example 3.4, we can retrieve
some well-known orders. For example, if fa1 ; a2 g ¼ f0; 1g and 2
fa3 ; a4 g ¼ f0; 1g, we obtain lexicographic orders. Moreover, all Proof. The linearity of IV
B is straight as LIV ð½0; 1Þ  ðLð½0; 1ÞÞ . In

these 4-admissible orders are particular examples of the construc- addition, it refines the partial order given by Atanassov due to
tion in Proposition 3.2, with c ¼ d ¼ 0 for A1 ; A2 and a ¼ b ¼ 0 for the fact that the order relation, B3 ;B4 , has been reversed.
A3 and A4 . In particular, if B1 ¼ B3 and B2 ¼ B4 , then B1 ;B2 ¼ B3 ;B4 and,
In [18] it was proven that given an a 2 ½0; 1Þ then all admissible consequently, the same order is used to compare both intervals
orders a;b on Lð½0; 1Þ with b > a coincide. Then, different aggrega- although in the second one the order is reversed. h
tion functions could generate the same admissible order. This also
affects to admissible orders generated as in Proposition 3.2. For Proposition 4.2. Let A ¼< A1 ; A2 ; A3 ; A4 > be four aggregation func-
instance, tions, Ai : ½0; 14 ! ½0; 1 such that for all ðp1 ; p2 ; p3 ; p4 Þ;

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
6 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx

Fig. 2. Visual representation of interval-valued intuitionistic pairs. The white zone represents the subset of LIV ð½0; 1Þ in which such pairs are allowed.

ðq1 ; q2 ; q3 ; q4 Þ 2 ½0; 14 the equalities Ai ðp1 ; p2 ; p3 ; p4 Þ ¼ Ai ðq1 ; q2 ; L ð½0; 1Þ ¼ fsjs ¼ ðs1 ; s2 Þ such that 0 6 s2 6 s1 6 1g:
q3 ; q4 Þ for all i 2 f1; . . . ; 4g only hold if ðp1 ; p2 ; p3 ; p4 Þ ¼ ðq1 ; q2 ; q3 ; q4 Þ. Then in Proposition 4.2 it is enough that to see, given ðp1 ; q1 Þ ¼
An IVAIF-admissible order IV A on LIV ð½0; 1Þ, is defined as follows: ð½p1 ; p1 ; ðq1 ; q1 ÞÞ; ðp2 ; q2 Þ ¼ ð½p2 ; p2 ; ðq2 ; q2 ÞÞ 2 Lð½0; 1Þ  L ð½0; 1Þ,the
ðx1 ; y1 Þ IV
A ðx2 ; y2 Þ if and only if one of the following (mutually equalities Ai ðp1 ; p1 ; q1 ; q1 Þ ¼ Ai ðp2 ; p2 ; q2 ; q2 Þ hold if and only if
exclusive) conditions is satisfied. ðp1 ; q1 Þ ¼ ðp2 ; q2 Þ.
However, in order to simplify notation we have imposed a
(i) A1 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ < A1 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ, slightly stronger restriction. Anyway, all the given examples in
(ii) A1 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ ¼ A1 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ and Section 3 satisfy it.
A2 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ < A2 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ,
Let a tuple A ¼ hA1 ; . . . ; A4 i of aggregation functions generate an
(iii) A1 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ ¼ A1 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ,
admissible order. Let Bi : ½0; 12 ! ½0; 1 be four aggregations such
A2 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ ¼ A2 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ, and
that
A3 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ < A3 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ,
(iv) A1 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ ¼ A1 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ,  Ai ðx; x; y; yÞ ¼ Bi ðx; xÞ for i 2 f1; 2g, and
A2 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ ¼ A2 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ,  Aj ðx; x; y; yÞ ¼ Bj ðy; yÞ for j 2 f3; 4g,
A3 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ ¼ A3 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ, and
A4 ðx1 ; x1 ; 1  y1 ; 1  y1 Þ 6 A4 ðx2 ; x2 ; 1  y2 ; 1  y2 Þ. Then the orders IV IV
A and B may be different. To guarantee that
they are actually different it is enough that B3 ðy1 ; y1 Þ < B3 ðy2 ; y2 Þ
Proof. The linearity is warranted because the equalities only hold and simultaneously B3 ð1  y1 ; 1  y1 Þ > B3 ð1  y2 ; 1  y2 Þ hold true
if ðx1 ; x1 ; 1  y1 ; 1  y1 Þ ¼ ðx2 ; x2 ; 1  y2 ; 1  y2 Þ. for some y1 ; y2 2 Lð½0; 1Þ.
To check the second condition (that of refining the partial order) For instance, let B3 ðy; yÞ ¼ yy. Here, we have that for
in the statement of Definition 4.1, notice that if y1 ¼ ½0:2; 0:2 and y2 ¼ ½0:1; 0:9
x1 6 x2 ; x1 6 x2 ; y1 P y2 ; and y1 P y2 :
B3 ð0:2; 0:2Þ ¼ 0:04 < 0:09 ¼ B3 ð0:1; 0:9Þ
then B3 ð0:8; 0:8Þ ¼ 0:64 > 0:09 ¼ B3 ð0:9; 0:1Þ:
x1 6 x2 ; x1 6 x2 ; 1  y1 6 1  y2 ; and 1  y1 6 1  y2 ;
Proposition 4.3. Let a1 ; a2 ; a3 ; a4 2 ½0; 1, with a1 – a2 and a3 – a4 .
so consequently Ai ðx1 ; x1 ; 1  y1 ; 1  y1 Þ 6 Ai ðx2 ; x2 ; 1  y2 ; 1  y2 Þ for If
all i 2 f1; . . . ; 4g.
From now on we name the order generated by four aggregation  Ai ðx; x; y; yÞ ¼ K ai ðx; xÞ for i 2 f1; 2g, and
functions (as in Proposition 4.2) 4-IVAIF-admissible order. h
 Aj ðx; x; y; yÞ ¼ K aj ðy; yÞ for j 2 f3; 4g,

Remark 5. Given y 2 Lð½0; 1Þ, it follows that ð1  y; 1  yÞ 2 then the tuple A ¼ hA1 ; . . . ; A4 i generates a 4-IVAIF admissible order
L ð½0; 1Þ, where that is equal to IV
B being B ¼ hKa1 ; Ka2 ; Ka3 ; Ka4 i.

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx 7

Proof. The fact that the aggregation functions satisfy the condi- (2.1) Aggregation phase. A collective preference structure is
tions to generate a 4-IVAIF order is a simple calculation. To prove built from the set of individual homogeneous prefer-
the equality between the two orders notice that in this case the ence structures.
conditions iÞ and iiÞ of the order IV A are exactly equal to (2.2) Exploitation phase. A given method is applied to the
x1 Ka1 ;Ka2 x2 . Then, it is enough to prove that for all c, collective preference structure to obtain a selection
of alternatives.
Kc ð1  a1 ; 1  b1 Þ < Kc ð1  a2 ; 1  b2 Þ
is equivalent to Kc ða1 ; b1 Þ > Kc ða2 ; b2 Þ. We use the theoretical developments in previous sections in the
But exploitation phase of the group decision making problem consid-
ered by Nayagam [23]. In particular, we consider the adaptation
Kc ð1  a1 ; 1  b1 Þ < Kc ð1  a2 ; 1  b2 Þ
of this problem done by Zhang et al. [24]. In this adptation, authors
() 1  a1 þ cð1  b1  ð1  a1 ÞÞ < 1  a2 þ cð1  b2  ð1  a2 ÞÞ
consider that there exists a panel with four possible alternatives for
() 1  a1 þ cða1  b1 Þ < 1  a2 þ cða2  b2 Þ
investment:
() a2  cða2  b2 Þ < a1  cða1  b1 Þ
() a2 þ cðb2  a2 Þ < a1 þ cðb1  a1 Þ
() Kc ða2 ; b2 Þ < Kc ða1 ; b1 Þ; (1) z1 is a car company,
(2) z2 is a food company,
so the proof is complete. h (3) z3 is a computer company,
(4) z4 is an arms company.
Example 4.1. Let IV A be the order generated by A ¼
< K0:25 ; K0:75 ; K0:25 ; K0:75 >. Consider the elements z1 ¼ ð½0:15; It is necessary to choose the best company for investment.
0:35; ½0:2; 0:5Þ and z2 ¼ ð½0:15; 0:35; ½0:1; 0:9Þ 2 LIV ð½0; 1Þ. Since Let the data in [24] be our collective preference matrix. In the
their membership degrees are identical we only need to compare exploitation phase we use the voting method which consists in
their nonmembership degrees. aggregating the values in each row of the collective matrix Rc in
In fact, such a way that, at the end, we have as many values (pairs of inter-
vals) as rows. Since these latter values are not comparable through
K0:25 ð0:2; 0:5Þ ¼ 0:2 þ 0:25 ð0:5  0:2Þ ¼ 0:275 < 0:3 the partial order, we will select the alternative associated to the
¼ 0:1 þ 0:25 ð0:9  0:1Þ ¼ K0:25 ð0:1; 0:9Þ largest pair, according to a considered linear order.
0 1
 ð½0:4; 0:5; ½0:3; 0:4Þ ð½0:4; 0:6; ½0:2; 0:4Þ ð½0:1; 0:3; ½0:5; 0:6Þ
and ð½0:15; 0:35; ½0:1; 0:9Þ IV
A ð½0:15; 0:35; ½0:2; 0:5Þ. B ð½0:6; 0:7; ½0:2; 0:3Þ
B  ð½0:6; 0:7; ½0:2; 0:3Þ ð½0:4; 0:8; ½0:1; 0:2Þ C
C
Rc ¼ B C:
@ ð½0:3; 0:6; ½0:3; 0:4Þ ð½0:5; 0:6; ½0:3; 0:4Þ  ð½0:4; 0:5; ½0:1; 0:3Þ A
ð½0:7; 0:8; ½0:1; 0:2Þ ð½0:6; 0:7; ½0:1; 0:3Þ ð½0:3; 0:4; ½0:1; 0:2Þ 
5. Application to decision making
To aggregate the values of each row of Rc we use the concept of
Decision making problems may be summarized as follows. We interval-valued intuitionistic t-norms.
have a set of p alternatives:
Z ¼ fz1 ; ; zp g Definition 5.1. A mapping T : ðLIV ð½0; 1ÞÞ2 ! LIV ð½0; 1Þ is an
interval-valued intuitionistic t-norm if it is symmetric, associative,
and a set of n > 2 experts: increasing with respect to the partial order  given by Atanassov
E ¼ fe1 ; ; en g: (also called monotone) and Tððx; yÞ; ð1; 0ÞÞ ¼ ðx; yÞ.

Each of the latter provides her/his preferences on the former It is easy to see that, if we take the classical product t-norm,
set of alternatives by means of a preference relation in the following T P ðx; yÞ ¼ x y, and its dual t-conorm with respect to the standard
way: negation, SP ðx; yÞ ¼ x þ y  x y, the following expression is an
0 1 interval-valued intuitionistic t-norm: Tððx; yÞ; ðz; tÞÞ ¼ ð½x z; x z;
 r ðelÞ12 r ðelÞ1p
½y þ t  y t; y þ t  y tÞ.
B r ðelÞ21  r ðelÞ2p C
B C Applying T to each row of Rc we get a new matrix, say Rg, given
rel ¼ B C: ð4Þ
@  A by:
r ðelÞp1  0 1
z1 ¼ ð½0:016; 0:090; ½0:720; 0:856Þ
Here r ðelÞij , with i – j, expresses to what extent the expert l (with B z ¼ ð½0:144; 0:392; ½0:424; 0:608Þ C
B 2 C
Rg ¼ B C:
l 2 f1; ; ng) prefers the alternative zi over the alternative zj . @ z3 ¼ ð½0:060; 0:180; ½0:559; 0:748Þ A
We must reach a decision of selecting either an alternative or a z4 ¼ ð½0:126; 0:224; ½0:271; 0:552Þ
set of alternatives, which is (are) optimal as regards the experts
assessments. In this setting, as regards the partial order , it follows
In [20], it is stated that the resolution of a group decision mak- z1  z3  z2 and z1  z3  z4 ;
ing problem consists of two steps:
but z2 and z4 are not comparable.
(1) Uniform representation of information. In this phase, the For this reason we consider the 4-IVAIF-admissible order IV
A
heterogeneous information for the problem (the information defined through the following aggregation functions.
can be represented by means of preference orderings or util-
2 2 8 8
ity functions or fuzzy preference relations) is translated into  A1 ðx1 ; x1 ; y1 ; y1 Þ ¼ 20 x1 þ 20 x1 þ 20 y1 þ 20 y1
homogeneous information by means of different trans-  A2 ðx1 ; x1 ; y1 ; y1 Þ ¼ 10 5
x þ 20
20 1
3
x1 þ 20 2
y1 þ 20 y1
formation functions (see [22]). 1
 A3 ðx1 ; x1 ; y1 ; y1 Þ ¼ 20 x1 þ 10 8
x þ 20 1
y1 þ 20 y1
20 1
(2) Application of a selection procedure. This procedure consists
 A4 ðx1 ; x1 ; y1 ; y1 Þ ¼ 14 x1 þ 14 x1 þ 14 y1 þ 14 y1 .
of two phases:

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
8 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx

With this order, we have z1 IV IV IV

A z3 A z2 A z4 and the selected
However, when the extension does not affect to the whole
firm is arms company. power set, but to some suitable superset (smaller than the power
However, as it happened in [18] different 4-IVAIFS-admissible set), perhaps it may still happen that an extension accomplishing
orders can lead to different rankings and hence the selection of the aprioristic criteria is possible, after all. As far as we know, an analy-
best alternative for a given decision making problem can be forced. sis of this kind where we start with the canonical order of the real
For instance, in our case if we take IV with line (instead of a linear order on a finite set), and try to extend it to
A
A ¼< K0:25 ; K0:75 ; K0:25 ; K0:75 > it comes out that the best alternative the set of closed intervals of real numbers, following some list of
is the second one, since z1 IV IV IV criteria that have been set beforehand, is an open problem.
A z3 A z4 A z2 . Nevertheless, for the
IV We leave for future works the interpretation of the length of the
order ð½x1 ; x1 ; ½y1 ; y1 Þ A ð½x2 ; x2 ; ½y2 ; y2 Þ if and only if
intervals in a given decision making problem and its relation with
ignorance functions and possibility theory.
 (x1 < x2 ), or
 (x1 ¼ x2 and x1 < x2 ), or
 (x1 ¼ x2 ; x1 ¼ x2 and y1 > y2 ), or else Acknowledgments
 (x1 ¼ x2 ; x1 ¼ x2 ; y1 ¼ y2 and y1 P y2 )
The work has been supported by projects TIN2013-40765-P and
we have that z1 IV IV IV
A z3 A z4 A z2 and the best alternative is MTM2012-37894-C02-02 of the Spanish Ministry of Science and
the second one. the Research Services of the Universidad Publica de Navarra. The
To cope with this situation the following algorithm takes differ- last two authors acknowledge the support of the grants VEGA
ent 4-IVAIFS-admissible orders into account simultaneously. 1/0419/13 and VEGA 1/0420/15.

(1) To select several linear orders built with the methods devel-
oped in the previous sections. References
(2) For each order, to apply in the exploitation phase the voting [1] E. Barrenechea, J. Fernandez, M. Pagola, F. Chiclana, H. Bustince, Construction
method with the same aggregations. For instance, of interval-valued fuzzy preference relations from ignorance functions and
T ¼ ðT P ; SP Þ. fuzzy preference relations. Application to decision making, Knowl.-Based Syst.
58 (2014) 33–44.
(3) To select the alternative which appears as the best placed in
[2] H. Bustince, E. Barrenechea, M. Pagola, J. Fernandez, Z.S. Xu, B. Bedregal, J.
the majority of all the so-obtained rankings. Montero, H. Hagras, F. Herrera, B. De Baets, A historical account of types of
fuzzy sets and their relationships, IEEE Trans. Fuzzy Syst. (2015) (submitted for
In our considered problem, the chosen alternative through this publication).
[3] J. Sanz, M. Galar, A. Jurio, A. Brugos, M. Pagola, H. Bustince, Medical diagnosis of
algorithm is the second one. That is, we must invest our money in a cardiovascular diseases using an interval-valued fuzzy rule-based
food company. Clearly, the nature of the problem will impose the classification system, Appl. Soft Comput. 20 (2014) 103–111.
number of linear orders to be considered and/or the conditions that [4] H. Bustince, E. Barrenechea, M. Pagola, Generation of interval-valued fuzzy and
Atanassov intuitionistic fuzzy connectives from fuzzy connectives and from K a
will force us to use alternative methods. operators: laws for conjunctions and disjunctions, amplitude, Int. J. Intell. Syst.
23 (6) (2008) 680–714.
[5] G. Deschrijver, C. Cornelis, E.E. Kerre, On the representation of intuitionistic
6. Conclusions fuzzy t-norms and t-conorms, IEEE Trans. Fuzzy Syst. 12 (1) (2004) 45–61.
[6] K.T. Atanassov, Intuitionistic fuzzy sets, Fuzzy Set Syst. 20 (1) (1986) 87–96.
In this work we have studied how to construct linear orders [7] K.T. Atanassov, G. Gargov, Interval-valued intuitionistic fuzzy sets, Fuzzy Set
Syst. 31 (3) (1989) 343–349.
between pairs of intervals on Lð½0; 1Þ that can be used to construct [8] H. Bustince, M. Pagola, E. Barrenechea, J. Fernandez, P. Melo-Pinto, P. Couto,
linear orders in Atanassov interval-valued intuitionistic fuzzy sets. H.R. Tizhoosh, J. Montero, Ignorance functions. An application to the
We have applied this operator to group decision making problems calculation of the threshold in prostate ultrasound images, Fuzzy Set Syst.
161 (1) (2010) 20–36.
giving two algorithms, the first one for a particular linear order and [9] T. Calvo, A. Kolesárová, M. Komorníková, R. Mesiar, Aggregation operators:
the second one which mixes different linear orders. properties, classes and construction methods, in: T. Calvo, A. Kolesárová, M.
As a possible development for future research, somewhat Komorníková, R. Mesiar (Eds.), Aggregation Operators, Studies in Fuzziness
and Soft Computing, vol. 97, Physica-verlag HD, 2002, pp. 3–104.
related to the main ideas introduced throughout the present
[10] G. Beliakov, H. Bustince, D.P. Goswami, U.K. Mukherjee, N.R. Pal, On averaging
manuscript, we point out the introduction of different orderings operators for Atanassov’s intuitionistic fuzzy sets, Inf. Sci. 181 (6) (2011)
on families of intervals of the real line could be also analyzed from 1116–1124.
the point of view of extensions of the canonical ordering of the real [11] G. Beliakov, A. Pradera, T. Calvo, Aggregation Functions: A Guide for
Practitioners, Studies in Fuzziness and Soft Computing, vol. 221, Springer,
line to a superset (namely, Lð½0; 1Þ) following a suitable set of cri- 2007.
teria established a priori. The real line can be immediately embed- [12] G. Birkhoff, Lattice Theory, American Mathematical Society, vol. 25,
ded into Lð½0; 1Þ by just considering each real number x as the Colloquium Publications, 1967.
[13] B.A. Davey, H.A. Priestley, Introduction to Lattices and Order, second ed.,
degenerate interval ½x; x. Cambridge University Press, Cambridge, 2002.
A similar typical problem corresponds to the extension of linear [14] M.J. Campión, J.C. Candeal, E. Induráin, Representability of binary relations
orders from a finite set to its power set. Indeed, although it is always through fuzzy numbers, Fuzzy Set Syst. 157 (1) (2006) 1–19.
[15] J.C. Candeal, J. Gutiérrez García, E. Induráin, Universal codomains to represent
possible to extend a linear order from a given finite set U to its interval orders, Int. J. Uncertain Fuzz. 17 (2) (2009) 197–219.
power set, a typical question that gave rise to some classical papers [16] G. Bosi, J. Gutiérrez García, E. Induráin, Unified representability of total
from the 1970s on (see e.g. [25–28]), is whether or not it is possible preorders and interval orders through a single function: the lattice approach,
Order 26 (3) (2009) 255–275.
to perform an extension that follows a list of criteria imposed a [17] G. Bosi, A numerical representation of semiorders on a countable set, Riv. Mat.
priori. Sometimes, the extension is not possible because the criteria Sci. Econ. Soc. 16 (2) (1993) 15–19.
used are, so-to-say, contradictory. But, in addition, there are other [18] H. Bustince, J. Fernandez, A. Kolesárová, R. Mesiar, Generation of linear orders
for intervals by means of aggregation functions, Fuzzy Set Syst. 220 (2013)
situations in which the extension is not possible because of a com-
69–77.
binatorial explosion which, due to the bigger cardinality of the [19] M. Komorníková, R. Mesiar, Aggregation functions on bounded partially
power set of U, does not leave room to rank all the terms of the ordered sets and their classification, Fuzzy Set Syst. 175 (1) (2011) 48–56.
power set in an extended linear order, accomplishing all the cri- [20] J. Fodor, M. Roubens, Fuzzy preference modelling and multicriteria decision
support, in: Theory and Decision Library, Kluwer Academic Publishers, 1994.
teria. Perhaps the most famous result in this direction is the so- [21] M. Grabisch, J.L. Marichal, R. Mesiar, E. Pap, Aggregation Functions, Cambridge
called Kannai–Peleg impossibility theorem (see [26]). University Press, Cambridge, 2009.

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004
 L. De Miguel et al. / Information Fusion xxx (2015) xxx–xxx 9

[22] F. Chiclana, F. Herrera, E. Herrera-Viedma, Integrating three representation [25] P. Gärdenfors, Manipulation of social choice functions, J. Econ. Theory 13 (2)
models in fuzzy multipurpose decision making based on fuzzy preference (1976) 217–228.
relations, Fuzzy Set Syst. 97 (1) (1998) 33–48. [26] Y. Kannai, B. Peleg, A note on the extension of an order on a set to the power
[23] V.L.G. Nayagam, S. Muralikrishnan, G. Sivaramanr, Multi-criteria decision- set, J. Econ. Theory 32 (1) (1984) 172–175.
making method based on interval-valued intuitionistic fuzzy sets, Expert Syst. [27] P.C. Fishburn, Comment on the Kannai–Peleg impossibility theorem for
Appl. 38 (2011) 464–467. extending orders, J. Econ. Theory 32 (1) (1984) 176–179.
[24] Q. Zhang, H. Xing, F. Liu, J. Ye, P. Tang, Some new entropy measures for interval- [28] W. Bossert, On the extension of preferences over a set to the power set: an
valued intuitionistic fuzzy sets based on distances and their relationships with axiomatic characterization of a quasi-ordering, J. Econ. Theory 49 (1) (1989)
similarity and inclusion measures, Inf. Sci. 283 (2014) 55–69. 84–92.

Please cite this article in press as: L. De Miguel et al., Construction of admissible linear orders for interval-valued Atanassov intuitionistic fuzzy sets with an
application to decision making, Informat. Fusion (2015), http://dx.doi.org/10.1016/j.inffus.2015.03.004