Research Article
A New Ranking Principle For Ordering Trapezoidal
Intuitionistic Fuzzy Numbers
Lakshmana Gomathi Nayagam Velu, Jeevaraj Selvaraj, and Dhanasekaran Ponnialagan
Department of Mathematics, National Institute of Technology, Tiruchirappalli, India Correspondence should be addressed to Jeevaraj Selvaraj; alba.jeevi@gmail.com
Received 27 June 2016; Revised 4 November 2016; Accepted 13 November 2016; Published 6 February 2017 Academic Editor: Jose Egea
Copyright Β© 2017 Lakshmana Gomathi Nayagam Velu et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Modelling real life (industrial) problems using intuitionistic fuzzy numbers is inevitable in the present scenario due to their efficiency in solving problems and their accuracy in the results. Particularly, trapezoidal intuitionistic fuzzy numbers (TrIFNs) are widely used in describing impreciseness and incompleteness of a data. Any intuitionistic fuzzy decision-making problem requires the ranking procedure for intuitionistic fuzzy numbers. Ranking trapezoidal intuitionistic fuzzy numbers play an important role in problems involving incomplete and uncertain information. The available intuitionistic fuzzy decision-making methods cannot perform well in all types of problems, due to the partial ordering on the set of intuitionistic fuzzy numbers. In this paper, a new total ordering on the class of TrIFNs using eight different score functions, namely, imprecise score, nonvague score, incomplete score, accuracy score, spread score, nonaccuracy score, left area score, and right area score, is achieved and our proposed method is validated using illustrative examples. Significance of our proposed method with familiar existing methods is discussed.
1. Introduction
Classical set theory cannot be the better choice for modelling problems involving qualitative or imprecise information. To model such problems, fuzzy number was introduced by Jain [1] and some operations on fuzzy numbers are defined in [2, 3]. Intuitionistic fuzzy numbers are comparatively better in modelling real life problems involving uncertainties and imprecise information. Particularly, trapezoidal intuitionistic fuzzy numbers are more effective in describing impreciseness and incompleteness of a data. To resolve the task of compar-ing trapezoidal intuitionistic fuzzy numbers, many authors have proposed different ranking methods but none of them yield a total order on the class of TrIFNs with finite number of scores. Different ranking methodologies on the class of intuitionistic fuzzy numbers are discussed in [4, 5].
Nehi and Maleki [6] generalised the idea of natural ordering on real numbers to the triangular intuitionistic fuzzy numbers (TIFNs) by adopting a statistical view point. Nehi [7] compared TIFNs using lexicographic technique. Li [2] developed the idea of value and ambiguity of a triangular intuitionistic fuzzy number and introduced the
new ranking method using the concept of the ratio of the value index to the ambiguity index. Ye [8] presented the new ranking method using expected value of a trapezoidal intuitionistic fuzzy number and solved the decision-making problem using weighted expected value of TrIFN. Dubey and Mehara [9] extended the concept of value and ambiguity to the slightly modified TIFN and proposed a new approach to solve intuitionistic fuzzy linear programing problem. Nehi [7] introduced the concept of characteristic values of membership and nonmembership functions of TrIFN and proposed a new ranking method for trapezoidal intuitionistic fuzzy numbers by using it. Zhang and Nan [10] developed a compromise ratio ranking method for fuzzy multiattribute decision-making (MADM) problem based on the concept that larger TIFN among other TIFNs will be closer to the maximum value index and it will be far away from the minimum ambiguity index simultaneously. Kumar and Kaur [11] proposed the ranking method for TrIFNs by modifying Nehiβs [10] method. Zeng et al. [12] introduced a new ranking method for TrIFNs by extending the concept value and ambiguity of TIFN defined in Li [2]. Wan and Dong [13] introduced the concept of lower and upper weighted
Volume 2017, Article ID 3049041, 24 pages https://doi.org/10.1155/2017/3049041
possibility mean and possibility mean for a trapezoidal intuitionistic fuzzy numbers and proposed the new ranking method by use of it. Different ranking methods and their applications on multicriteria decision-making problem and other domains are studied in ([14β20]). Lakshmana Gomathi Nayagam et al. [19, 21, 22] have introduced a complete ranking procedure on the class of intuitionistic fuzzy numbers using countable number of parameter. In this paper, a new total ordering on the class of TrIFNs using finite number of score functions is achieved. Also the limitations and drawbacks of all the abovementioned methods are discussed and the efficiency of our proposed method is shown by comparing all existing methods.
This paper is organised in the following manner. After introduction, some important definitions on intuitionistic fuzzy numbers are given in Section 2. The different subclasses of TrIFNs are introduced and the new score functions on these subclasses are established in Section 3. In Section 4, a complete ranking on the class of trapezoidal intuitionistic fuzzy numbers by using score functions defined in Section 3 is explained. The ranking procedure is explained in detail with several examples and also our proposed method is compared with some other existing methods in the Section 5. Finally conclusions are given in Section 6.
2. Preliminaries
Here we give a brief review of some preliminaries.
Definition 1 (Atanassov [23]). Letπ be a nonempty set. An
intuitionistic fuzzy set (IFS) π΄ in π is defined by π΄ = {β¨π₯, ππ΄(π₯), ]π΄(π₯)β© | π₯ β π}, where ππ΄(π₯) : π β [0, 1] and ]π΄(π₯) : π β [0, 1], π₯ β π with the conditions 0 β€ ππ΄(π₯) + ]π΄(π₯) β€ 1, βπ₯ β π. The numbers ππ΄(π₯), ]π΄(π₯) β [0, 1] denote the degree of membership and nonmembership ofπ₯ to lie inπ΄, respectively. For each intuitionistic fuzzy subset π΄ inπ, ππ΄(π₯) = 1 β ππ΄(π₯) β ]π΄(π₯) is called hesitancy degree of π₯ to lie in π΄.
Definition 2 (Burillo et al., [24]). An intuitionistic fuzzy
numberπ΄ = (ππ΄, ]π΄) in the set of real numbers π is defined as ππ΄(π₯) = { { { { { { { { { { { { { { { ππ΄(π₯) if π β€ π₯ β€ π1 1 if π1β€ π₯ β€ π2 ππ΄(π₯) if π2β€ π₯ β€ π 0 Otherwise, ]π΄(π₯) = { { { { { { { { { { { { { { { βπ΄(π₯) if π β€ π₯ β€ π1 0 if π1β€ π₯ β€ π2 ππ΄(π₯) if π2β€ π₯ β€ π 1 Otherwise, (1) where0 β€ ππ΄(π₯) + ]π΄(π₯) β€ 1 and π, π1, π2, π, π, π1, π2, π β R such thatπ β€ π, π1 β€ π1 β€ π2 β€ π2, π β€ π, and four functions ππ΄, ππ΄, βπ΄, ππ΄ : R β [0, 1] are the legs of membership
(0, 1) (0, 0) (e, 1) (a, 1) (b1, 1) (b2, 1) (c, 1)(g, 1) πc A πA gA βA ]A fA kA (a ,0 ) (f1 ,0 ) (b1 ,0 ) (f2 , 0) (b2 ,0 ) (c ,0 )
Figure 1: Intuitionistic fuzzy number.
functionππ΄and nonmembership function]π΄. The functions ππ΄andππ΄are nondecreasing continuous functions and the functionsβπ΄andππ΄are nonincreasing continuous functions. An intuitionistic fuzzy number{(π, π1, π2, π), (π, π1, π2, π)} with(π, π1, π2, π) β€ (π, π1, π2, π)πis shown in Figure 1.
Definition 3. A trapezoidal intuitionistic fuzzy number π΄
with parametersπ β€ π, π1 β€ π1β€ π2β€ π2, π β€ π is denoted as π΄ = {(π, π1, π2, π), (π, π1, π2, π)} in the set of real numbers R is an intuitionistic fuzzy number whose membership function and nonmembership function are given as
ππ΄(π₯) = { { { { { { { { { { { { { { { { { π₯ β π1 π2β π1 if π1β€ π₯ β€ π2 1 if π2β€ π₯ β€ π3 π4β π₯ π4β π3 if π3β€ π₯ β€ π4 0 Otherwise, ]π΄(π₯) = { { { { { { { { { { { { { { { { { π₯ β π2 π1β π2 if π1β€ π₯ β€ π2 0 if π2β€ π₯ β€ π3 π₯ β π3 π4β π3 if π3β€ π₯ β€ π4 1 Otherwise. (2)
Ifπ2 = π3(andπ2 = π3) in a trapezoidal intuitionistic fuzzy numberπ΄, we have the triangular intuitionistic fuzzy num-bers as special case of the trapezoidal intuitionistic fuzzy numbers.
A trapezoidal intuitionistic fuzzy number π΄ = {(π, π1, π2, π), (π, π1, π2, π)} with π1 β€ π1,π2 β₯ π2,π β€ π, and π β₯ π is shown in Figure 2.
We note that the condition(π, π1, π2, π) β€ (π, π1, π2, π)π of the trapezoidal intuitionistic fuzzy numberπ΄ = {(π, π1, π2, π), (π, π1, π2, π)} whose membership and nonmembership fuzzy numbers ofπ΄ are (π, π1, π2, π) and (π, π1, π2, π) implies
(0, 1) (0, 0) (e, 1) (a, 1) (b1, 1) (b2, 1) (c, 1) (g, 1) πcA πA ]A (a ,0 ) (f1 ,0 ) (b1 ,0 ) (f2 , 0) (b2 ,0 ) (c ,0 )
Figure 2: Trapezoidal intuitionistic fuzzy number.
π1 β€ π1,π2 β₯ π2,π β€ π, and π β₯ π on the legs of trapezoidal intuitionistic fuzzy number.
Definition 4 (Atanassov and Gargov, [25]). Letπ·[0, 1] be the
set of all closed subintervals of the interval[0, 1]. An interval valued intuitionistic fuzzy set on a setπ ΜΈ= π is an expression given byπ΄ = {β¨π₯, ππ΄(π₯), ]π΄(π₯)β© : π₯ β π}, where ππ΄ : π β π·[0, 1], ]π΄: π β π·[0, 1] with the condition 0 < supπ₯ππ΄(π₯)+ supπ₯]π΄(π₯) β€ 1.
The intervals ππ΄(π₯) and ]π΄(π₯) denote, respectively, the degree of belongingness and nonbelongingness of the ele-mentπ₯ to the set π΄. Thus for each π₯ β π, ππ΄(π₯) and ]π΄(π₯) are closed intervals whose lower and upper end points are, respectively, denoted byππ΄πΏ(π₯), ππ΄π(π₯) and ]π΄πΏ(π₯), ππ΄π(π₯). We denoteπ΄ = {β¨π₯, [ππ΄πΏ(π₯), ππ΄π(π₯)], []π΄πΏ(π₯), ]π΄π(π₯)]β© : π₯ β π}, where 0 < ππ΄(π₯) + ]π΄(π₯) β€ 1.
For each elementπ₯ β π, we can compute the unknown degree (hesitance degree) of belongingness ππ΄(π₯) to π΄ as ππ΄(π₯) = 1 β ππ΄(π₯) β ]π΄(π₯) = [1 β ππ΄π(π₯) β ]π΄π(π₯), 1 β ππ΄πΏ(π₯) β ]π΄πΏ(π₯)]. An intuitionistic fuzzy interval number (IFIN) is denoted byπ΄ = ([π, π], [π, π]) for convenience.
Definition 5. LetπΌ, π½β (0, 1]. An (πΌ, π½)-cut of a trapezoidal
intuitionistic fuzzy number, denoted by(πΌ,π½)π΄πΌ, is defined as
(πΌ,π½)π΄ πΌ = {(πΌππ΄πΌ, π½] π΄πΌ)}, where πΌπ π΄πΌand π½]
π΄πΌ areπΌ-cut and π½-cut of membership and nonmembership trapezoidal fuzzy numbers.
The(πΌ, π½)-cut of trapezoidal intuitionistic fuzzy number π΄πΌ = β¨(π1π, π2π, π3π, π4π), (πσΈ 1], πσΈ 2], π3]σΈ , π4]σΈ )β© is given by (πΌ,π½)π΄ πΌ = {[πΌππ΄πΌπΏ,πΌππ΄πΌπ], [π½]π΄πΌπΏ,π½]π΄πΌπ]} = ([π1π + πΌ(π2π β π1π), π4πβ πΌ(π4πβ π3π)], [πσΈ 1]+ π½ (π2]σΈ β πσΈ 1]), π4]σΈ β π½ (π4]σΈ β πσΈ 3])]), whereπΌππ΄ πΌπΏ and πΌπ
π΄πΌπ represents the lower and upper end points of theπΌ-cut of the membership function of π΄πΌ and
π½]
π΄πΌπΏ andπ½]π΄πΌπrepresent the lower and upper end points of theπ½-cut of the nonmembership function of π΄πΌ, respectively.
3. Different Classes of Trapezoidal
Intuitionistic Fuzzy Numbers
In this section the entire class of TrIFNs is partitioned into eight subclasses and further score functions using different concepts are defined in order to give total order on the entire class of TrIFNs and some theorems related to these concepts are established.
In this paper, π΄πΌ and π΅πΌ always denote trapezoidal intuitionistic fuzzy number (TrIFN).
π΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© with condi-tions thatπσΈ 1]β€ π1πandπ2]σΈ β€ π2π,π3]σΈ β₯ π3πandπ4]σΈ β₯ π4πand π΅πΌ = β¨(π1π, π2π, π3π, π4π), (πσΈ
1], π2]σΈ , π3]σΈ , π4]σΈ )β© with the condition thatπ1]σΈ β€ π1πandπ2]σΈ β€ π2π,π3]σΈ β₯ π3π, andπ4]σΈ β₯ π4πunless otherwise specifically stated.
3.1. Imprecise Score of a Trapezoidal Intuitionistic Fuzzy Number. In this subsection, a new subclass of TrIFNs is
introduced. The imprecise score function on this class of trapezoidal intuitionistic fuzzy numbers is defined and some of its properties are studied using illustrative examples.
Definition 6. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β©
andπ΅πΌ= β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , π3]σΈ , π4]σΈ )β© be two TrIFNs. A special subclassπΆ1of the set of TrIFNs consist of TrIFNs for which every pair ofπ΄πΌandπ΅πΌare withπ1π β₯ π1π, π2π β₯ π2π, π3π β€ π3π, π4π β€ π4π and π1]σΈ β₯ π1]σΈ , π2]σΈ β₯ π2]σΈ , π3]σΈ β€ πσΈ
3], π4]σΈ β€ π4]σΈ . The imprecise relation onπΆ1is denoted asβ and defined as follows: ifπ΄πΌ, π΅πΌ β πΆ1 β TrIFN such that π΄πΌ β π΅πΌthenπ1π β₯ π1π, π2π β₯ π2π, π3π β€ π3π, π4π β€ π4πand πσΈ
1]β₯ π1]σΈ , πσΈ 2]β₯ π2]σΈ , π3]σΈ β€ π3]σΈ , π4]σΈ β€ π4]σΈ .
Ifπ΄πΌ β π΅πΌthen one of the above inequalities becomes strict inequality.
Note 1. By Definition 6, we note that any pair of members of
πΆ1are related underβ.
The score function which measures the preciseness is defined as follows.
Definition 7. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β©
be a trapezoidal intuitionistic fuzzy number with π1π, π2π, π3π, π4π, πσΈ
1], πσΈ 2], π3]σΈ , πσΈ 4]β [0, 1]. Then the imprecise score of a trapezoidal intuitionistic fuzzy numberπ΄πΌis defined by
π½1(π΄πΌ) = β«1 0 [πΌπ π΄πΌπΏβπΌππ΄πΌπ] 4 ππΌ + β« 1 0 [π½] π΄πΌπΏβπ½]π΄πΌπ] 4 ππ½ = β«1 0 (π1π+ πΌ (π2πβ π1π)) 4 ππΌ β β«1 0 (π4πβ πΌ (π4πβ π3π)) 4 ππΌ
+ β«1 0 (π1]σΈ + π½ (π2]σΈ β π1]σΈ )) 4 ππ½ β β«1 0 (πσΈ 4]β π½ (π4]σΈ β π3]σΈ )) 4 ππ½ =(π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ π2]σΈ ) β (π3]σΈ + πσΈ 4]) 8 . (3) The proofs of the following propositions are immediate from the above definition and hence they are omitted.
Proposition 8. For any real number π β [0, 1], π½1(π) = 0.
Proposition 9. If π΄πΌ = (π1π, π2π, π3π, π4π) is a trapezoidal
fuzzy number, thenπ½1(π΄πΌ) = ((π1π+ π2π) β (π3π+ π4π))/4.
Proposition 10. Let π΄πΌ = β¨(π1π, π2π, π3π), (π1]σΈ , π2]σΈ , πσΈ 3])β© be a
triangular intuitionistic fuzzy number. Thenπ½1(π΄πΌ) = ((π1πβ
π3π) + (π1]σΈ β π3]σΈ ))/4.
Proposition 11. Let π΄πΌ= ([π1π, π2π], [πσΈ
1], π2]σΈ ]) be an interval
valued intuitionistic fuzzy number. Thenπ½1(π΄πΌ) = ((π1πβπ2π)+
(πσΈ
1]β π2]σΈ ))/2.
Theorem 12. Let π΄πΌandπ΅πΌ β πΆ1. Ifπ΄πΌβ π΅πΌthenπ½1(π΄πΌ) >
π½1(π΅πΌ).
Proof. Letπ΄πΌβ π΅πΌ. We claim thatπ½1(π΄πΌ) β π½1(π΅πΌ) > 0. By Definition 6, we have π1πβ₯ π1π, π2πβ₯ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β₯ π1]σΈ , πσΈ 2]β₯ π2]σΈ , πσΈ 3]β€ π3]σΈ , πσΈ 4]β€ π4]σΈ . (4) Now, 8 (π½1(π΄πΌ) β π½1(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(π1]σΈ β π1]σΈ ) + (π2]σΈ β π2]σΈ ) + (π3]σΈ β πσΈ 3]) + (π4]σΈ β πσΈ 4])] . (5)
From (4) it is very easy to see that all the terms in (5) are pos-itive. Therefore their sum is also pospos-itive. From Definition 6, we know that at least one of the above inequalities in (4) becomes strict inequality and hence we getπ½1(π΄πΌ) β π½1(π΅πΌ) > 0.
Theorem 13. Let π΄πΌandπ΅πΌ β πΆ1such thatπ½1(π΄πΌ) = π½1(π΅πΌ);
thenπ΄πΌ= π΅πΌ.
Proof. Letπ½1(π΄πΌ) = π½1(π΅πΌ). We claim that π΄πΌ= π΅πΌ. By Definition 6, without loss of generality, we have
π1πβ₯ π1π, π2πβ₯ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β₯ π1]σΈ , πσΈ 2]β₯ π2]σΈ , πσΈ 3]β€ π3]σΈ , πσΈ 4]β€ π4]σΈ . (6) Now, 8 (π½1(π΄πΌ) β π½1(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(πσΈ 1]β π1]σΈ ) + (π2]σΈ β π2]σΈ ) + (π3]σΈ β π3]σΈ ) + (π4]σΈ β πσΈ 4])] = 0. (7)
Therefore from (6) and (7), it is clear that all the terms in (7) are positive and their sum gives zero only when each term is equal to zero. Henceπ1π = π1π,π2π = π2π,π3π = π3π,π4π = π4πandπ1]σΈ = π1]σΈ , π2]σΈ = π2]σΈ , π3]σΈ = π3]σΈ , πσΈ 4] = π4]σΈ , hence the proof.
Note 2. The imprecise score can be calculated to any TrIFN
butπ½1gives total order inπΆ1.
Definition 14. Ifπ½1(π΄πΌ) > π½1(π΅πΌ)(π½1(π΄πΌ) < π½1(π΅πΌ)), then π΄πΌ>
π΅πΌ(π΄πΌ< π΅πΌ).
The following example explains the ranking procedure introduced in Definition 14. Example 15. Letπ΄πΌ= β¨(0.1, 0.2, 0.35, 0.5), (0, 0.15, 0.45, 0.6)β© and π΅πΌ = β¨(0.15, 0.2, 0.35, 0.45), (0.1, 0.1, 0.5, 0.55)β© β πΆ1. Now π½1(π΄πΌ) = β0.18125 and π½1(π΅πΌ) = β0.1625, π½1(π΄πΌ) < π½1(π΅πΌ). Henceπ΄πΌ< π΅πΌ. Example 16. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© andπ΅πΌ = β¨(π1π+ π, π2π+ π, π3πβ π, π4πβ π), (πσΈ 1]β π, π2]σΈ β π, π3]σΈ + π, π4]σΈ + π)β© β TrIFN but not in πΆ1, whereπ β [0, 1].
Nowπ½1(π΄πΌ) = π½1(π΅πΌ) = ((π1π+ π2π) β (π3π+ π4π) + (π1]σΈ + πσΈ
2]) β (π3]σΈ + π4]σΈ ))/8. But π΄πΌ ΜΈ= π΅πΌ. Pictorial representation of this example is given in Figure 3.
Example 16 shows thatπ½1is not enough cover the entire class of TrIFNs. Therefore it is needed for us to define another
(0, 1) (0, 0) bσ³° 1] aσ³°1] b σ³° 2] σ³° 2a] a2π b2π a1π b1π a3π b3π a4π b4π aσ³° 4] a σ³° 3] bσ³° 4] b σ³° 3] Figure 3: Pictorial representation of Example 16.
score function that can cover some subclass of TrIFNs which cannot be covered byπ½1.
3.2. Nonvague Score of a Trapezoidal Intuitionistic Fuzzy Number. In this subsection, another subclass of TrIFNs is
introduced using nonvague relation. The nonvague score function on this class of trapezoidal intuitionistic fuzzy numbers is defined and some of its properties are studied using illustrative examples.
Definition 17. Letπ΄πΌandπ΅πΌbe two TrIFNs.
A special subclassπΆ2of the set of TrIFNs consist of TrIFNs for which every pair ofπ΄πΌandπ΅πΌare withπ1π β₯ π1π, π2π β₯ π2π, π3π β€ π3π, π4π β€ π4π and π1]σΈ β€ π1]σΈ , π2]σΈ β€ π2]σΈ , π3]σΈ β₯ πσΈ
3], π4]σΈ β₯ π4]σΈ . The nonvague relation onπΆ2is denoted asβ and defined as follows: ifπ΄πΌ, π΅πΌ β πΆ2 β TrIFN such that π΄πΌ β π΅πΌthenπ1π β₯ π1π, π2π β₯ π2π, π3π β€ π3π, π4π β€ π4πand πσΈ
1]β€ π1]σΈ , πσΈ 2]β€ π2]σΈ , π3]σΈ β₯ π3]σΈ , π4]σΈ β₯ π4]σΈ .
Ifπ΄πΌ β π΅πΌthen one of the above inequalities becomes strict inequality.
Note 3. By Definition 17, we note that any pair of members of
πΆ2are related underβ.
The score function which measures the nonvagueness is defined as follows.
Definition 18. Let π΄πΌ be a trapezoidal intuitionistic fuzzy number. Then the nonvague score of a trapezoidal intuition-istic fuzzy numberπ΄πΌis defined by
π½2(π΄πΌ) = β«1 0 [πΌπ π΄πΌπΏβπΌππ΄πΌπ] 4 ππΌ β β« 1 0 [π½] π΄πΌπΏβπ½]π΄πΌπ] 4 ππ½ =(π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ π2]σΈ ) + (π3]σΈ + πσΈ 4]) 8 . (8)
The proofs of the following propositions are immediate from Definition 18 and hence they are omitted.
Proposition 19. For any real number π β [0, 1], π½2(π) = 0.
Proposition 20. If π΄πΌ = (π1π, π2π, π3π, π4π) is a trapezoidal
fuzzy number, thenπ½2(π΄πΌ) = 0.
Proposition 21. Let π΄πΌ= β¨(π1π, π2π, π3π), (πσΈ
1], π2]σΈ , πσΈ 3])β© be a
triangular intuitionistic fuzzy number. Thenπ½2(π΄πΌ) = ((π1πβ
π3π) + (βπσΈ
1]+ π3]σΈ ))/8.
Proposition 22. Let π΄πΌ= ([π1π, π2π], [πσΈ
1], π2]σΈ ]) be an interval
valued intuitionistic fuzzy number. Thenπ½2(π΄πΌ) = ((π1πβπ2π)+
(βπσΈ 1]+ πσΈ 2]))/4.
Theorem 23. Let π΄πΌandπ΅πΌ β πΆ2. Ifπ΄πΌβ π΅πΌthenπ½2(π΄πΌ) >
π½2(π΅πΌ).
Proof. Let us assumeπ΄πΌβ π΅πΌ. We claim thatπ½2(π΄πΌ)βπ½2(π΅πΌ) > 0. By Definition 17, we have π1πβ₯ π1π, π2πβ₯ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β€ π1]σΈ , πσΈ 2]β€ π2]σΈ , πσΈ 3]β₯ π3]σΈ , πσΈ 4]β₯ π4]σΈ . (9) Now, 8 (π½2(π΄πΌ) β π½2(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(π1]σΈ β π1]σΈ ) + (π2]σΈ β πσΈ 2]) + (πσΈ 3]β π3]σΈ ) + (πσΈ 4]β π4]σΈ )] . (10)
From (9) it is very easy to see that all the terms in (10) are pos-itive. Therefore their sum is also pospos-itive. From Definition 17, we know that at least one of the above inequalities in (9) becomes strict inequality and hence we getπ½2(π΄πΌ) β π½2(π΅πΌ) > 0.
Theorem 24. Let π΄πΌandπ΅πΌ β πΆ2. Ifπ½2(π΄πΌ) = π½2(π΅πΌ) then
π΄πΌ= π΅πΌ.
By Definition 17, without loss of generality, we have π1πβ₯ π1π, π2πβ₯ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β€ π1]σΈ , πσΈ 2]β€ π2]σΈ , πσΈ 3]β₯ π3]σΈ , πσΈ 4]β₯ π4]σΈ . (11) Now, 8 (π½2(π΄πΌ) β π½2(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(π1]σΈ β π1]σΈ ) + (π2]σΈ β πσΈ 2]) + (π3]σΈ β π3]σΈ ) + (πσΈ 4]β π4]σΈ )] = 0. (12)
Therefore from (11) and (12), it is clear that all the terms in (12) are positive; therefore their sum gives zero only when each term is equal to zero. Henceπ1π = π1π, π2π = π2π, π3π = π3π, π4π = π4πandπ1]σΈ = π1]σΈ , πσΈ 2] = π2]σΈ , π3]σΈ = π3]σΈ , π4]σΈ = π4]σΈ . Henceπ΄πΌ= π΅πΌ.
Note 4. The nonvague score can be calculated to any TrIFN.
Butπ½2gives total order onπΆ2.
Definition 25. Ifπ½2(π΄πΌ) > π½2(π΅πΌ)(π½2(π΄πΌ) < π½2(π΅πΌ)), then π΄πΌ>
π΅πΌ(π΄πΌ< π΅πΌ).
Ranking relation defined above is explained in the follow-ing example. Example 26. Letπ΄πΌ = β¨(0.15, 0.25, 0.35, 0.45), (0.1, 0.2, 0.4, 0.5)β© and π΅πΌ = β¨(0.18, 0.25, 0.35, 0.4), (0, 0.15, 0.45, 0.55)β© β πΆ2. Nowπ½2(π΄πΌ) = 0.025 and π½2(π΅πΌ) = 0.06625, π½2(π΅πΌ) > π½2(π΄πΌ). Henceπ΅πΌ> π΄πΌ.
Example 27 shows the inefficiency ofπ½1in comparing any two arbitrary TrIFNs and the importance of defining new score functionπ½2. Example 27. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© andπ΅πΌ = β¨(π1π+ π, π2π+ π, π3πβ π, π4πβ π), (πσΈ 1]β π, π2]σΈ β π, πσΈ 3]+ π, π4]σΈ + π)β© β TrIFN and π β [0, 1]. Now π½1(π΄πΌ) = π½1(π΅πΌ) =(π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ π2]σΈ ) β (πσΈ 3]+ πσΈ 4]) 8 σ³¨β π΄πΌ= π΅πΌ. (13) (0, 1) (0, 0) b1σ³°] a1σ³°] b σ³° 2] a σ³° 2] a2π b2π a1π b1π a3πb3π a4π b4π a4σ³°] a σ³° 3] b4σ³°] b σ³° 3] Figure 4: Pictorial representation of Example 28.
But π½2(π΄πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ πσΈ 2]) + (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ πσΈ 2]) + (πσΈ 3]+ π4]σΈ ) 8 + π. (14) Thereforeπ½2(π΅πΌ) > π½2(π΄πΌ); hence π΅πΌ> π΄πΌ. Example 28. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© andπ΅πΌ = β¨(π1π+ π, π2π+ π, π3π+ π, π4π+ π), (πσΈ 1]+ π/2, π2]σΈ + π/2, πσΈ
3]+π/2, π4]σΈ +π/2)β© β TrIFN and π β [0, 1] but not in πΆ2. Now π½1(π΄πΌ) = π½1(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ πσΈ 2]) β (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΄πΌ) = π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ πσΈ 2]) + (πσΈ 3]+ π4]σΈ ) 8 . (15) Butπ΄πΌ ΜΈ= π΅πΌ.
Pictorial representation of this example is given in Fig-ure 4.
From Example 28 it is easy to see thatπ½1andπ½2are not enough to cover the entire class of TrIFNs. Therefore in the next subsection we are introducing a new score function which covers some more subclasses of TrIFNs that cannot be covered byπ½1andπ½2.
3.3. Incomplete Score of a Trapezoidal Intuitionistic Fuzzy Number. In this subsection, another class of TrIFNs is
trapezoidal intuitionistic fuzzy numbers is defined and some of its properties are studied using illustrative examples.
Definition 29. Letπ΄πΌandπ΅πΌbe two TrIFNs.
A special subclassπΆ3of the set of TrIFNs consist of TrIFNs for which every pair ofπ΄πΌandπ΅πΌare withπ1π β€ π1π, π2π β€ π2π, π3π β€ π3π, π4π β€ π4π and π1]σΈ β₯ π1]σΈ , π2]σΈ β₯ π2]σΈ , π3]σΈ β₯ πσΈ
3], π4]σΈ β₯ π4]σΈ . The incomplete relation onπΆ3 is denoted as β and defined as follows: if π΄πΌ, π΅πΌ β πΆ3 β TrIFN such that π΄πΌ β π΅πΌthenπ1π β€ π1π, π2π β€ π2π, π3π β€ π3π, π4π β€ π4πand πσΈ
1]β₯ π1]σΈ , πσΈ 2]β₯ π2]σΈ , π3]σΈ β₯ π3]σΈ , π4]σΈ β₯ π4]σΈ .
Ifπ΄πΌ β π΅πΌthen one of the above inequalities becomes strict inequality.
Note 5. By Definition 29, we note that any pair of members
ofπΆ3are related underβ.
The incomplete score function which measures the com-pleteness is defined as follows.
Definition 30. Let π΄πΌ be a trapezoidal intuitionistic fuzzy number. Then the incomplete score ofπ΄πΌis defined by
π½3(π΄πΌ) = β«1 0 β [πΌππ΄πΌπΏ +πΌππ΄πΌπ] 4 ππΌ + β«1 0 [π½] π΄πΌπΏ+π½]π΄πΌπ] 4 ππ½ = (π σΈ 1]+ π2]σΈ + π3]σΈ + π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 . (16)
The proofs of the following propositions are immediate from the above definition and hence they are omitted.
Proposition 31. For any real number π β [0, 1], π½3(π) = 0.
Proposition 32. If π΄πΌ = (π1π, π2π, π3π, π4π) is a trapezoidal
fuzzy number, thenπ½3(π΄πΌ) = 0.
Proposition 33. Let π΄πΌ= β¨(π1π, π2π, π3π), (πσΈ
1], πσΈ 2], π3]σΈ )β© be a
triangular intuitionistic fuzzy number. Then
π½3(π΄πΌ) = (π σΈ 1]+ 2πσΈ 2]+ πσΈ 3]) β (π1π+ 2π2π+ π3π) 8 . (17) Proposition 34. Let π΄πΌ= ([π1π, π2π], [πσΈ 1], πσΈ 2]]) be an interval
valued intuitionistic fuzzy number. Then
π½3(π΄πΌ) = (π σΈ
1]+ π2]σΈ ) β (π1π+ π2π)
4 . (18)
Theorem 35. Let π΄πΌ, π΅πΌ β πΆ3. If π΄πΌ β π΅πΌthen π½3(π΄πΌ) >
π½3(π΅πΌ).
Proof. Let us assume thatπ΄πΌ β π΅πΌ. We claim thatπ½3(π΄πΌ) β π½3(π΅πΌ) > 0. By Definition 29, we have π1πβ€ π1π, π2πβ€ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β₯ π1]σΈ , πσΈ 2]β₯ π2]σΈ , πσΈ 3]β₯ π3]σΈ , πσΈ 4]β₯ π4]σΈ . (19) Now, 8 (π½3(π΄πΌ) β π½3(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(πσΈ 1]β π1]σΈ ) + (π2]σΈ β π2]σΈ ) + (πσΈ 3]β π3]σΈ ) + (πσΈ 4]β π4]σΈ )] . (20)
From (19) it is very easy to see that all the terms in (20) are positive. Therefore their sum is also positive. From Defini-tion 29, we know that at least one of the above inequalities in (19) becomes strict inequality and hence we getπ½3(π΄πΌ) β π½3(π΅πΌ) > 0, hence the proof.
Theorem 36. Let π΄πΌ, π΅πΌ β πΆ3. Ifπ½3(π΄πΌ) = π½3(π΅πΌ) then π΄πΌ =
π΅πΌ.
Proof. Let us assumeπ½3(π΄πΌ) = π½2(π΅πΌ). We prove that π΄πΌ= π΅πΌ. By Definition 29, without loss of generality, we have
π1πβ€ π1π, π2πβ€ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β₯ π1]σΈ , πσΈ 2]β₯ π2]σΈ , πσΈ 3]β₯ π3]σΈ , πσΈ 4]β₯ π4]σΈ . (21) Now, 8 (π½3(π΄πΌ) β π½3(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(πσΈ 1]β π1]σΈ ) + (π2]σΈ β π2]σΈ ) + (π3]σΈ β π3]σΈ ) + (π4]σΈ β π4]σΈ )] = 0. (22)
Therefore from (21) and (22), it is clear that all the terms in (22) are positive and therefore their sum gives zero only when each term is equal to zero. Henceπ1π = π1π, π2π = π2π, π3π = π3π, π4π = π4πandπ1]σΈ = π1]σΈ , πσΈ 2] = π2]σΈ , π3]σΈ = π3]σΈ , π4]σΈ = π4]σΈ . Henceπ΄πΌ= π΅πΌ.
Note 6. The incomplete score can be calculated to any TrIFN.
Butπ½3gives total order onπΆ3.
Definition 37. Ifπ½3(π΄πΌ) > π½3(π΅πΌ)(π½3(π΄πΌ) < π½3(π΅πΌ)), then π΄πΌ>
π΅πΌ(π΄πΌ< π΅πΌ).
The following example is used to explain the ranking procedure defined in Definition 37.
Example 38. Letπ΄πΌ = β¨(0.15, 0.3, 0.4, 0.45), (0.1, 0.25, 0.45,
0.6)β© and π΅πΌ = β¨(0.2, 0.4, 0.45, 0.55), (0, 0.15, 0.45, 0.55)β© β πΆ3. Nowπ½3(π΄πΌ) = 0.0125 and π½3(π΅πΌ) = β0.05625, π½3(π΄πΌ) > π½3(π΅πΌ).
Henceπ΄πΌ> π΅πΌ.
The inefficiency of the score functions π½1 and π½2 in discriminating any two arbitrary TrIFNs and the ability of π½3in comparing arbitrary TrIFNs is shown in the following example. Example 39. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© andπ΅πΌ= β¨(π1π+ π, π2π+ π, π3π+ π, π4π+ π), (πσΈ 1]+ π/2, π2]σΈ + π/2, πσΈ 3]+ π/2, πσΈ 4]+ π/2)β© β TrIFN and π β [0, 1]. Now π½1(π΄πΌ) = π½1(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ πσΈ 2]) β (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΄πΌ) = π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ πσΈ 2]) + (πσΈ 3]+ π4]σΈ ) 8 . (23) But π½3(π΄πΌ) =(π σΈ 1]+ π2]σΈ + π3]σΈ + π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 , π½3(π΅πΌ) =(π σΈ 1]+ π2]σΈ + π3]σΈ + π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 +π4 σ³¨β π½3(π΅πΌ) > π½3(π΄πΌ) . (24) Henceπ΅πΌ> π΄πΌ.
The inefficiency of the score functionsπ½1toπ½3in the task of comparing TrIFNs is explained in the following example.
(0, 1) (0, 0) b1σ³°] a1σ³°] b σ³° 2] a σ³° 2] a2π b2π a1π b1π a3π b3π a4π b4π a4σ³°] a σ³° 3] b4σ³°] b σ³° 3] Figure 5: Pictorial representation of Example 40.
Example 40. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β©
andπ΅πΌ= β¨(π1π+π, π2π+π, π3π+π, π4π+π), (π1]σΈ +π, πσΈ 2]+π, π3]σΈ + π, πσΈ
4]+ π)β© β TrIFN and π β [0, 1] but not in πΆ3. Now π½1(π΄πΌ) = π½1(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ π2]σΈ ) β (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΄πΌ) = π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ π2]σΈ ) + (πσΈ 3]+ π4]σΈ ) 8 , π½3(π΄πΌ) = π½3(π΅πΌ) = (π σΈ 1]+ π2]σΈ + πσΈ 3]+ π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 . (25) Butπ΄πΌ ΜΈ= π΅πΌ.
Pictorial representation of this example is given in Fig-ure 5.
Example 40 shows thatπ½1, π½2, andπ½3cannot be sufficient to cover the entire class of TrIFNs and the class of TrIFNs in the above example excite us to define new score function that can fill the subclass of TrIFNs which cannot be covered byπ½1, π½2, andπ½3.
3.4. Accuracy Score of a Trapezoidal Intuitionistic Fuzzy Number. In this subsection, a new subclass of TrIFNs using
accuracy relation is introduced and the accuracy score func-tion on this class of trapezoidal intuifunc-tionistic fuzzy numbers is defined and some of its properties are studied using illustrative examples.
Definition 41. Letπ΄πΌandπ΅πΌbe two TrIFNs.
A special subclassπΆ4of the set of TrIFNs consist of TrIFNs for which every pair ofπ΄πΌandπ΅πΌare withπ1π β€ π1π, π2π β€ π2π, π3π β€ π3π, π4π β€ π4π and π1]σΈ β€ π1]σΈ , π2]σΈ β€ π2]σΈ , π3]σΈ β€ πσΈ
3], π4]σΈ β€ π4]σΈ . The accuracy relation onπΆ4is denoted asβͺ― and defined as follows: ifπ΄πΌ, π΅πΌ β πΆ4 β TrIFN such that π΄πΌ βͺ― π΅πΌthenπ1π β€ π1π, π2π β€ π2π, π3π β€ π3π, π4π β€ π4πand πσΈ
1]β€ π1]σΈ , πσΈ 2]β€ π2]σΈ , π3]σΈ β€ π3]σΈ , π4]σΈ β€ π4]σΈ .
Ifπ΄πΌ βΊ π΅πΌthen one of the above inequalities becomes strict inequality.
Note 7. By Definition 41, we note that any pair of members of
πΆ4are related underβΊ.
The score function which measures the accuracy is defined as follows.
Definition 42. Let π΄πΌ be a trapezoidal intuitionistic fuzzy number. Then the accuracy score ofπ΄πΌis defined by
π½4(π΄πΌ) = β«1 0 [πΌπ π΄πΌπΏ +πΌππ΄πΌπ] 4 ππΌ + β« 1 0 [π½] π΄πΌπΏ +π½]π΄πΌπ] 4 ππ½ = (π1π+ π2π+ π3π+ π4π) + (π σΈ 1]+ π2]σΈ + πσΈ 3]+ πσΈ 4]) 8 . (26)
The proofs of the following propositions are immediate from Definition 42 and hence they are omitted.
Proposition 43. For any real number π β [0, 1], π½4(π) = π.
Proposition 44. For any trapezoidal fuzzy number π΄πΌ =
(π1π, π2π, π3π, π4π), π½4(π΄πΌ) = (π1π+ π2π+ π3π+ π4π)/4.
Proposition 45. Let π΄πΌ = β¨(π1π, π2π, π3π), (πσΈ
1], πσΈ 2], π3]σΈ )β© be a
triangular intuitionistic fuzzy number. Thenπ½4(π΄πΌ) = ((π1π+
2π2π+ π3π) + (πσΈ
1]+ 2π2]σΈ + π3]σΈ ))/8.
Proposition 46. Let π΄πΌ= ([π1π, π2π], [πσΈ 1], π2]σΈ ]) be an interval
valued intuitionistic fuzzy number. Thenπ½4(π΄πΌ) = ((π1π+π2π)+
(π1]σΈ + π2]σΈ ))/4.
Note 8. The accuracy score can be calculated to any TrIFN.
Butπ½4gives total order onπΆ4which is proved in Theorems 47 and 48.
Theorem 47. Let π΄πΌ, π΅πΌ β πΆ4. Ifπ΄πΌ βΊ π΅πΌ thenπ½4(π΄πΌ) <
π½4(π΅πΌ).
Proof. Let us assume thatπ΄πΌ βΊ π΅πΌ. We claim thatπ½4(π΅πΌ) β π½4(π΄πΌ) > 0. By Definition 41, we have π1πβ€ π1π, π2πβ€ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β€ π1]σΈ , πσΈ 2]β€ π2]σΈ , πσΈ 3]β€ π3]σΈ , πσΈ 4]β€ π4]σΈ . (27) Now, 8 (π½4(π΅πΌ) β π½4(π΄πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(π1]σΈ β π1]σΈ ) + (πσΈ 2]β πσΈ 2]) + (π3]σΈ β πσΈ 3]) + (π4]σΈ β πσΈ 4])] . (28)
From (27) it is very easy to see that all the terms in (28) are positive. Therefore their sum is also positive. From Defini-tion 41, we know that at least one of the above inequalities in (27) becomes strict inequality and hence we getπ½4(π΅πΌ) β π½4(π΄πΌ) > 0.
Theorem 48. Let π΄πΌ, π΅πΌβ πΆ4such thatπ½4(π΄πΌ) = π½4(π΅πΌ); then
π΄πΌ= π΅πΌ.
Proof. Let us assumeπ½4(π΄πΌ) = π½4(π΅πΌ). We claim that π΄πΌ= π΅πΌ. By Definition 41, without loss of generality, we have
π1πβ€ π1π, π2πβ€ π2π, π3πβ€ π3π, π4πβ€ π4π, πσΈ 1]β₯ π1]σΈ , πσΈ 2]β₯ π2]σΈ , πσΈ 3]β₯ π3]σΈ , πσΈ 4]β₯ π4]σΈ . (29) Now, 8 (π½4(π΅πΌ) β π½4(π΄πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(π1]σΈ β π1]σΈ ) + (π2]σΈ β πσΈ 2]) + (π3]σΈ β πσΈ 3]) + (π4]σΈ β π4]σΈ )] = 0. (30)
Therefore from (29) and (30), it is clear that all the terms in (30) are positive and therefore their sum gives zero only when each term is equal to zero. Henceπ1π = π1π, π2π = π2π, π3π = π3π, π4π = π4πandπσΈ 1] = π1]σΈ ,π2]σΈ = π2]σΈ , πσΈ 3] = π3]σΈ , πσΈ 4] = π4]σΈ . Henceπ΄πΌ= π΅πΌ.
Definition 49. Ifπ½4(π΄πΌ) > π½4(π΅πΌ)(π½4(π΄πΌ) < π½4(π΅πΌ)) then π΄πΌ>
π΅πΌ(π΄πΌ< π΅πΌ).
Ranking procedure defined in Definition 49 is illustrated in the following example.
Example 50. Let π΄πΌ = β¨(0.1, 0.25, 0.37, 0.49), (0.1, 0.2, 0.4,
0.6)β© and π΅πΌ = β¨(0.2, 0.37, 0.47, 0.6), (0.15, 0.32, 0.52, 0.7)β© β πΆ4. Nowπ½4(π΄πΌ) = 0.31375 and π½4(π΅πΌ) = 0.41625, π½4(π΄πΌ) < π½4(π΅πΌ).
Henceπ΄πΌ< π΅πΌ.
The relative importance of the score function π½4 in discriminating two different TrIFNs whenπ½1, π½2, andπ½3fail to discriminate them is explained by Example 51.
Example 51. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© andπ΅πΌ = β¨(π1π+ π, π2π+ π, π3π+ π, π4π+ π), (πσΈ 1]+ π, π2]σΈ + π, πσΈ 3]+ π, π4]σΈ + π)β© β TrIFN, and π β [0, 1]. Now π½1(π΄πΌ) = π½1(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ π2]σΈ ) β (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΄πΌ) = π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ π2]σΈ ) + (πσΈ 3]+ π4]σΈ ) 8 , π½3(π΄πΌ) = π½3(π΅πΌ) = (π σΈ 1]+ π2]σΈ + π3]σΈ + π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 . (31) But π½4(π΄πΌ) =(π1π+ π2π+ π3π+ π4π) + (π σΈ 1]+ π2]σΈ + π3]σΈ + πσΈ 4]) 8 , π½4(π΅πΌ) =(π1π+ π2π+ π3π+ π4π) + (π σΈ 1]+ π2]σΈ + π3]σΈ + πσΈ 4]) 8 + π σ³¨β π½4(π΄πΌ) < π½4(π΅πΌ) . (32) Henceπ΄πΌ< π΅πΌ. Example 52. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© andπ΅πΌ = β¨(π1π+ π, π2πβ π, π3π+ π, π4πβ π), (πσΈ 1]+ π, π2]σΈ β π, π3]σΈ + π, π4]σΈ β π)β© β TrIFN and π β [0, 1] but not in πΆ4.
(0, 1) (0, 0) b σ³° 1] a σ³° 1] b σ³° 2] σ³° 2a] a2π b2π a1π b1π a3π b3π a4π b4π a σ³° 4] a σ³° 3] b σ³° 4] b σ³° 3] Figure 6: Pictorial representation of Example 52.
Now π½1(π΄πΌ) = π½1(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ π2]σΈ ) β (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΄πΌ) = π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ π2]σΈ ) + (πσΈ 3]+ π4]σΈ ) 8 , π½3(π΄πΌ) = π½3(π΅πΌ) = (π σΈ 1]+ π2]σΈ + πσΈ 3]+ π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 , π½4(π΄πΌ) = π½4(π΅πΌ) = (π1π+ π2π+ π3π+ π4π) + (π σΈ 1]+ π2]σΈ + πσΈ 3]+ πσΈ 4]) 8 . (33) Butπ΄πΌ ΜΈ= π΅πΌ.
Pictorial representation of this example is given in Fig-ure 6.
Example 52 shows that even π½1, π½2, π½3, andπ½4 altogether are not enough to cover the entire class of TrIFNs; the class of TrIFNs in the above example excite us to define new score function that can cover the class of TrIFNs which cannot be covered byπ½1, π½2, π½3, andπ½4.
3.5. Spread Score of a Trapezoidal Intuitionistic Fuzzy Number.
In this subsection, another subclass of TrIFNs is introduced. The spread score function on this class of trapezoidal intu-itionistic fuzzy numbers is defined and some of its properties are studied using illustrative examples.
Definition 53. Letπ΄πΌandπ΅πΌbe two TrIFNs.
A special subclassπΆ5of the set of TrIFNs consist of TrIFNs for which every pair ofπ΄πΌandπ΅πΌare withπ1π β₯ π1π, π2π β€ π2π, π3π β₯ π3π, π4π β€ π4π and π1]σΈ β₯ π1]σΈ , π2]σΈ β€ π2]σΈ , π3]σΈ β₯ π3]σΈ , π4]σΈ β€ π4]σΈ . The preference relation onπΆ5 is denoted as
β΄ and defined as follows: if π΄πΌ, π΅πΌ β πΆ5 β TrIFN such that π΄πΌ β΄ π΅πΌthenπ1π β₯ π1π, π2π β€ π2π, π3π β₯ π3π, π4π β€ π4πand πσΈ
1]β₯ π1]σΈ , πσΈ 2]β€ π2]σΈ , π3]σΈ β₯ π3]σΈ , π4]σΈ β€ π4]σΈ . Ifπ΄πΌβ² π΅πΌthen one of the above inequalities becomes strict inequality.
Note 9. By Definition 53, we note that any pair of members
ofπΆ5are related underβ².
The score function which measures the spread is defined as follows.
Definition 54. Let π΄πΌ be a trapezoidal intuitionistic fuzzy number. Then the spread score ofπ΄πΌis defined by
π½5(π΄πΌ) = (π1π+ π3π) β (π2π+ π4π) β (π σΈ 2]+ πσΈ 4]) + (πσΈ 1]+ π3]σΈ ) 8 . (34)
The proofs of the following propositions are immediate from Definition 54 and hence they are omitted.
Proposition 55. For any real number π β [0, 1], π½5(π) = 0.
Proposition 56. If π΄πΌ = (π1π, π2π, π3π, π4π) is a trapezoidal
fuzzy number, thenπ½5(π΄πΌ) = ((π1π+ π3π) β (π2π+ π4π))/4.
Proposition 57. Let π΄πΌ= β¨(π1π, π2π, π3π), (πσΈ
1], πσΈ 2], π3]σΈ )β© be a
triangular intuitionistic fuzzy number. Thenπ½5(π΄πΌ) = (β(π3πβ
π1π) β (πσΈ
3]β π1]σΈ ))/8.
Proposition 58. Let π΄πΌ= ([π1π, π2π], [π1]σΈ , πσΈ 2]]) be an interval
valued intuitionistic fuzzy number. Thenπ½5(π΄πΌ) = 0.
Theorem 59. Let π΄πΌ, π΅πΌ β πΆ5. Ifπ΄πΌ β² π΅πΌ thenπ½5(π΄πΌ) >
π½5(π΅πΌ).
Proof. Let us assume thatπ΄πΌ β² π΅πΌ. We claim thatπ½5(π΄πΌ) β π½5(π΅πΌ) > 0. By Definition 53, we have π1πβ₯ π1π, π2πβ€ π2π, π3πβ₯ π3π, π4πβ€ π4π, πσΈ 1]β₯ π1]σΈ , πσΈ 2]β€ π2]σΈ , πσΈ 3]β₯ π3]σΈ , πσΈ 4]β€ π4]σΈ . (35) Now, 8 (π½5(π΄πΌ) β π½5(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(πσΈ 1]β π1]σΈ ) + (π2]σΈ β πσΈ 2]) + (πσΈ 3]β π3]σΈ ) + (π4]σΈ β π4]σΈ )] . (36)
From (35) it is very easy to see that all the terms in (36) are positive. Therefore their sum is also positive. From Defini-tion 53, we know that at least one of the above inequalities in (35) becomes strict inequality and hence we getπ½5(π΄πΌ) β π½5(π΅πΌ) > 0.
Theorem 60. Let π΄πΌ, π΅πΌ β πΆ5. Ifπ½5(π΄πΌ) = π½5(π΅πΌ) then π΄πΌ=
π΅πΌ.
Proof. Let us assumeπ½5(π΄πΌ) = π½5(π΅πΌ). We claim that π΄πΌ= π΅πΌ. By Definition 53, without loss of generality, we have
π1πβ₯ π1π, π2πβ€ π2π, π3πβ₯ π3π, π4πβ€ π4π, πσΈ 1]β₯ π1]σΈ , πσΈ 2]β€ π2]σΈ , πσΈ 3]β₯ π3]σΈ , πσΈ 4]β€ π4]σΈ . (37) Now, 8 (π½5(π΄πΌ) β π½5(π΅πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(πσΈ 1]β π1]σΈ ) + (π2]σΈ β πσΈ 2]) + (π3]σΈ β π3]σΈ ) + (π4]σΈ β πσΈ 4])] = 0. (38)
Therefore from (37) and (38), it is clear that all the terms in (38) are positive and therefore their sum gives zero only when each term is equal to zero. Henceπ1π = π1π, π2π = π2π, π3π = π3π, π4π = π4πandπ1]σΈ = π1]σΈ , πσΈ 2] = π2]σΈ , π3]σΈ = π3]σΈ , π4]σΈ = π4]σΈ . Henceπ΄πΌ= π΅πΌ.
Note 10. The spread score can be calculated to any TrIFN. But
π½5gives total order onπΆ5which is seen from Theorems 59 and 60.
Definition 61. Ifπ½5(π΄πΌ) > π½5(π΅πΌ)(π½5(π΄πΌ) < π½5(π΅πΌ)) then π΄πΌ>
π΅πΌ(π΄πΌ< π΅πΌ).
The following example is used to explain the ranking procedure defined onπΆ5.
Example 62. Let π΄πΌ = β¨(0.15, 0.23, 0.31, 0.45), (0.1, 0.16, 0.35, 0.5)β© and π΅πΌ = β¨(0.1, 0.25, 0.25, 0.5), (0, 0.2, 0.3, 0.6)β© β πΆ5. Nowπ½5(π΄πΌ) = β0.05375 and π½5(π΅πΌ) = β0.1125, π½5(π΄πΌ) > π½5(π΅πΌ). Henceπ΄πΌ> π΅πΌ. Example 63. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β© andπ΅πΌ = β¨(π1π+ π, π2πβ π, π3π+ π, π4πβ π), (πσΈ 1]+ π, π2]σΈ β π, πσΈ 3]+ π, π4]σΈ β π)β© β TrIFN and π β [0, 1]. Now π½1(π΄πΌ) = π½1(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ πσΈ 2]) β (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΄πΌ) = π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ πσΈ 2]) + (πσΈ 3]+ π4]σΈ ) 8 , π½3(π΄πΌ) = π½3(π΅πΌ) = (π σΈ 1]+ π2]σΈ + πσΈ 3]+ π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 , π½4(π΄πΌ) = π½4(π΅πΌ) = (π1π+ π2π+ π3π+ π4π) + (π σΈ 1]+ π2]σΈ + πσΈ 3]+ πσΈ 4]) 8 . (39) But π½5(π΄πΌ) =(π1π+ π3π) β (π2π+ π4π) + (π σΈ 1]+ πσΈ 3]) β (π2]σΈ + π4]σΈ ) 8 , π½5(π΅πΌ) =(π1π+ π3π) β (π2π+ π4π) + (π σΈ 1]+ πσΈ 3]) β (π2]σΈ + π4]σΈ ) 8 + π σ³¨β π½5(π΅πΌ) > π½5(π΄πΌ) . (40)
Hence π΅πΌ > π΄πΌ. In this example the importance of π½5 in ranking arbitrary TrIFNs is shown.
The inability ofπ½5in comparing any two TrIFNs is shown in the following example.
Example 64. Letπ΄πΌ = β¨(π1π, π2π, π3π, π4π), (π1]σΈ , π2]σΈ , πσΈ 3], π4]σΈ )β©
andπ΅πΌ= β¨(π1π+π, π2πβπ, π3π+π, π4πβπ), (π1]σΈ βπ, πσΈ 2]+π, π3]σΈ β π, π4]σΈ + π)β© β TrIFN and π β [0, 1] but not in πΆ5.
(0, 1) (0, 0) b σ³° 1] σ³° 1a] b σ³° 2] a σ³° 2] a2π b2π a1π b1π a3π b3π a4π b4π a σ³° 4] a σ³° 3] b σ³° 4] b σ³° 3] Figure 7: Pictorial representation of Example 64.
Now π½1(π΄πΌ) = π½1(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) + (π σΈ 1]+ πσΈ 2]) β (πσΈ 3]+ π4]σΈ ) 8 , π½2(π΄πΌ) = π½2(π΅πΌ) = (π1π+ π2π) β (π3π+ π4π) β (π σΈ 1]+ πσΈ 2]) + (πσΈ 3]+ π4]σΈ ) 8 , π½3(π΄πΌ) = π½3(π΅πΌ) = (π σΈ 1]+ π2]σΈ + π3]σΈ + π4]σΈ ) β (π1π+ π2π+ π3π+ π4π) 8 , π½4(π΄πΌ) = π½4(π΅πΌ) = (π1π+ π2π+ π3π+ π4π) + (π σΈ 1]+ πσΈ 2]+ π3]σΈ + π4]σΈ ) 8 , π½5(π΄πΌ) = π½5(π΅πΌ) = (π1π+ π3π) β (π2π+ π4π) + (π σΈ 1]+ πσΈ 3]) β (πσΈ 2]+ π4]σΈ ) 8 . (41) Butπ΄πΌ ΜΈ= π΅πΌ.
Pictorial representation of this example is given in Fig-ure 7.
Example 64 shows that all the above defined scores are not enough to cover the entire class of TrIFNs; therefore we are introducing another score function in the next subsection.
3.6. Nonaccuracy Score of a Trapezoidal Intuitionistic Fuzzy Number. In this subsection, a new subclass of TrIFNs is
introduced. The nonaccuracy score function on this class of trapezoidal intuitionistic fuzzy numbers is defined and some of its properties are studied using illustrative examples.
Definition 65. Letπ΄πΌandπ΅πΌbe two TrIFNs.
A special subclassπΆ6of the set of TrIFNs consist of TrIFNs for which every pair ofπ΄πΌandπ΅πΌare withπ1π β₯ π1π, π2π β€ π2π, π3π β₯ π3π, π4π β€ π4π and π1]σΈ β€ π1]σΈ , π2]σΈ β₯ π2]σΈ , π3]σΈ β€ πσΈ
3], π4]σΈ β₯ π4]σΈ . The exact relation onπΆ6is denoted asβ΅ and defined as follows: ifπ΄πΌ, π΅πΌβ πΆ6β TrIFN such that π΄πΌβ΅ π΅πΌ thenπ1π β₯ π1π, π2π β€ π2π, π3π β₯ π3π, π4π β€ π4πand πσΈ 1] β€ πσΈ
1], π2]σΈ β₯ π2]σΈ , π3]σΈ β€ π3]σΈ , π4]σΈ β₯ π4]σΈ . Ifπ΄πΌβ³ π΅πΌthen one of the above inequalities becomes strict inequality.
Note 11. By Definition 65, we note that any pair of members
ofπΆ6are related underβ³.
The score function which measures the nonaccuracy is defined as follows.
Definition 66. Let π΄πΌ be a trapezoidal intuitionistic fuzzy number. Then the nonaccuracy score ofπ΄πΌis defined by
π½6(π΄πΌ) = (π2π+ π4π) β (π1π+ π3π) + (π σΈ 1]+ πσΈ 3]) β (πσΈ 2]+ π4]σΈ ) 8 . (42)
The proofs of the following propositions are easy and hence they are omitted.
Proposition 67. For any real number π β [0, 1], π½6(π) = 0.
Proposition 68. If π΄πΌ = (π1π, π2π, π3π, π4π) is a trapezoidal
fuzzy number, thenπ½6(π΄πΌ) = 0.
Proposition 69. Let π΄πΌ= β¨(π1π, π2π, π3π), (πσΈ
1], π2]σΈ , π3]σΈ )β© be a
triangular intuitionistic fuzzy number. Thenπ½6(π΄πΌ) = ((π1]σΈ β
πσΈ
3]) β (π1πβ π3π))/8.
Proposition 70. Let π΄πΌ= ([π1π, π2π], [πσΈ
1], πσΈ 2]]) be an interval
valued intuitionistic fuzzy number. Thenπ½6(π΄πΌ) = 0.
Note 12. The nonaccuracy score can be calculated to any
TrIFN. But π½6 gives total order on πΆ6 which is proved in Theorems 71 and 72.
Theorem 71. Let π΄πΌ, π΅πΌ β πΆ6. Ifπ΄πΌ β³ π΅πΌthenπ½6(π΄πΌ) <
π½6(π΅πΌ).
Proof. Let us assume thatπ΄πΌ β³ π΅πΌ. We claim thatπ½6(π΅πΌ) β π½6(π΄πΌ) > 0. By Definition 65, we have π1πβ₯ π1π, π2πβ€ π2π, π3πβ₯ π3π, π4πβ€ π4π, πσΈ 1]β€ π1]σΈ , πσΈ 2]β₯ π2]σΈ , πσΈ 3]β€ π3]σΈ , πσΈ 4]β₯ π4]σΈ (43) and at least one of these inequalities becomes strict.
Now,
8 (π½6(π΅πΌ) β π½6(π΄πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(π1]σΈ β π1]σΈ ) + (πσΈ 2]β π2]σΈ ) + (π3]σΈ β π3]σΈ ) + (π4]σΈ β π4]σΈ )] .
(44)
From (43) it is very easy to see that all the terms in (44) are positive and therefore their sum is also positive. From Definition 65, we know that at least one of the above inequalities in (43) becomes strict inequality and hence we getπ½6(π΅πΌ) β π½6(π΄πΌ) > 0.
Theorem 72. Let π΄πΌ, π΅πΌ β πΆ6. Ifπ½6(π΄πΌ) = π½6(π΅πΌ) then π΄πΌ =
π΅πΌ.
Proof. Let us assumeπ½6(π΄πΌ) = π½6(π΅πΌ). We claim that π΄πΌ= π΅πΌ. By Definition 65, without loss of generality, we have
π1πβ₯ π1π, π2πβ€ π2π, π3πβ₯ π3π, π4πβ€ π4π, πσΈ 1]β€ π1]σΈ , πσΈ 2]β₯ π2]σΈ , πσΈ 3]β€ π3]σΈ , πσΈ 4]β₯ π4]σΈ . (45) Now, 8 (π½6(π΅πΌ) β π½6(π΄πΌ)) = [(π1πβ π1π) + (π2πβ π2π) + (π3πβ π3π) + (π4πβ π4π)] + [(π1]σΈ β π1]σΈ ) + (πσΈ 2]β π2]σΈ ) + (π3]σΈ β π3]σΈ ) + (π4]σΈ β π4]σΈ )] = 0. (46)
Therefore from (45) and (46), it is clear that all the terms in (46) are positive and therefore their sum gives zero only when each term is equal to zero. Henceπ1π = π1π, π2π = π2π, π3π = π3π, π4π = π4πandπ1]σΈ = π1]σΈ , πσΈ 2] = π2]σΈ , π3]σΈ = π3]σΈ , π4]σΈ = π4]σΈ . Henceπ΄πΌ= π΅πΌ.
Definition 73. Ifπ½6(π΄πΌ) > π½6(π΅πΌ)(π½6(π΄πΌ) < π½6(π΅πΌ)), then π΄πΌ>
π΅πΌ(π΄πΌ< π΅πΌ).
Ranking procedure introduced in Definition 73 is explained in Example 74.