On the first moments and semi-moments of fuzzy

variables based on a new measure and application for

portfolio selection with fuzzy returns
Justin Dzuche, Christian Deffo Tassak, Jules Sadefo-Kamdem, Louis Aimé
Fono

To cite this version:

Justin Dzuche, Christian Deffo Tassak, Jules Sadefo-Kamdem, Louis Aimé Fono. On the first moments
and semi-moments of fuzzy variables based on a new measure and application for portfolio selection
with fuzzy returns. [Research Report] WP MRE 2019.8, MRE - Montpellier Recherche en Economie.
2019. �hal-02433463�

HAL Id: hal-02433463

https://hal.umontpellier.fr/hal-02433463
Submitted on 9 Jan 2020

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 Justin DZUCHE, Christian DEFFO TASSAK,
Jules SADEFO KAMDEM et Louis Aimé FONO

« On the first moments and semi-moments of fuzzy

variables based on a new measure and application
for portfolio selection with fuzzy returns»

WP MRE 2019.8

Montpellier Recherche en Economie EA 7491 – Faculté d’Economie

Université de Montpellier - MUSE « Montpellier Université d’Excellence »
Contact : [email protected]
 On the first moments and semi-moments of fuzzy
variables based on a new measure and application
for portfolio selection with fuzzy returns
Justin Dzuchea , Christian Deffo Tassaka , Jules Sadefo Kamdemb ,
Louis Aimé Fonoa∗
a
Laboratory of Mathématics - Faculty of Sciences
a
University of Douala, B.P. 24157 Douala-Cameroon
b
MRE EA 7491 (Université de Montpellier, France) and
DFR SJE (Université de Guyane)
B.P. 792, 97337 Cayenne cedex, France

†
Corresponding author : [email protected] ; [email protected]

June 16, 2019

Abstract
Possibility, Necessity and Credibility measures are used in the liter-
ature in order to deal with imprecision. Recently, Yang and Iwamura
[11] introduced a new measure as convex linear combination of possi-
bility and necessity measures and they determined some of its axioms.
In this paper, we introduce characteristics (parameters) of a fuzzy vari-
able based on that measure, namely, Expected value, Variance, Semi-
Variance, Skewness, Kurtosis and Semi-Kurtosis. We determine some
properties of these characteristics and we compute them for trapezoidal
and triangular fuzzy variables. We display their application for the de-
termination of optimal portfolios when assets returns are described by
triangular or trapezoidal fuzzy variables.

Keywords: Fuzzy measure; Fuzzy variable; Expected value; Variance and

Semi-Variance; Skewness; Kurtosis and Semi-Kurtosis; Optimal portfolios.
∗
The work has been done under the research Grant No 17-497RG/M AT HS/AF/ACG −
F R3240297728 offered by The World Academy of Sciences (TWAS) to the Applied Math-
ematics to Social Sciences Research Group of the Laboratory of Mathematics-University
of Douala-Cameroon. The authors (members of the Laboratory of Mathematics) sincerely
thanks TWAS.

1
1 Introduction

1 Introduction
Uncertainty can be viewed as an aspect of randomness or ambiguity in real
life phenomena. Thereby, one needs a measure on fuzzy events (fuzzy mea-
sure) to analyze questions and problems dealing with such uncertainty. After
Zadeh’s work [12] introducing possibility and necessity measures, a vast liter-
ature appeared on fuzzy theory and fuzzy logic based on the aforementioned
measures. Twenty years later, Liu [6] introduced credibility measure as the
arithmetic mean of the two first measures. This measure has been used to
address other questions dealing with uncertainty (see Liu [7], Li et al. [5],
Huang et al. [4], Sadefo et al. [9] and Tassak et al. [10]). More recently,
Yang and Iwamura [11] introduced a new measure, denoted mλ , as a convex
linear combination of possibility and necessity measures (the weight λ of the
possibility measure in the combination is a real parameter in the interval
[0; 1]) and they determined some of its axioms. This new measure general-
ized the three previous ones and has been used in fuzzy chance-constrained
programming (Yang and Kakuzo [11], Dai et al. [1]) as a more generalized
approach compared with credibility constrained programming ones to find
optimal strategies for carbon capture. We used it in our recent conference
paper to determine Expected value and Variance of a fuzzy variable (Dzuche
et al. [2]).

This paper focuses on the recent measure for two major findings that
might be useful. First, the weight λ can be considered as the the decision
maker confidence on the fuzzy event degree of realization such that degree
1 means a complete confidence and degree 0 means no confidence. The
mλ − measure can become either possibility measure,or necessity measure or
credibility measure respectively for the decision marker who is self confident,
unconfident and neutral (the weight is equal to 12 ). It is a comprise of the
two first measures and this finding is similar to the idea of Hurwicz decision
criteria in Microeconomics which is a compromise of maximax and maximin
criteria. On the other hand, due to the fact that the new measure generalizes
the three previous ones in technical point of view, many theoretical results
already solved with one of the three first measures can be generalized with
the new measure. In addition, some specific results will be outlined.
The modest contribution of this paper is to study, by means of mλ , fuzzy
variables characteristics and to implement obtained results for portfolio op-
timization models with fuzzy returns. More specifically, in the continuation
of our recent work (Dzuche et al. [2]), we propose some new axioms of
mλ , and we introduce fuzzy variables characteristics w.r.t. to this measure.
We determine four first moments (expected value, variance, skewness and
kurtosis) and two first semi-moments (semi-variance and semi-kurtosis)of
trapezoidal and triangular fuzzy variables. We establish some properties
of these characteristics and the obtained theoretical results are applied in

2
2 Preliminaries

portfolio selection with fuzzy returns.

The paper is organized as follows. Section 2 recalls useful notions on
fuzzy sets, on four previous usual fuzzy measures. We display some new
axioms of the more recent one. In Section 3, we introduce and study fuzzy
variables characteristics. More precisely, we focus on the determination of
four first moments and two first semi-moments of trapezoidal and triangu-
lar fuzzy variables and we establish some properties of these characteristics.
Section 4 displays an application of these characteristics in optimal portfo-
lios selection where assets returns proposed by Huang [4] are described by
triangular fuzzy variables. We also apply these characteristics to find opti-
mal portfolios by using Tokyo stock exchange data provided by Hasuike et
al. [3]. Appendix contains proofs of some results.

2 Preliminaries
Throughout this paper, X is a nonempty set, namely, the universal set. A
fuzzy subset A of X is defined by its membership function: µA : X → [0, 1]
such that, to each x ∈ X, is associated µA (x) representing the membership
grade of x to A. A is denoted by {(x, µA (x)), x ∈ X}. If ∀x ∈ A, µA (x) ∈
{0, 1}, then A becomes a crisp subset of X.

Let ξ be a mapping from X to R described by its membership function µ

interpreted as: for any x ∈ R, µ(x) represents the degree that ξ takes value
x. ξ is said to be a fuzzy variable if ξ is measurable. A fuzzy variable ξ is
normal if ∃x0 ∈ R, µ(x0 ) = 1.
For ξ and η be two fuzzy variables and α a real number, their sum ξ + η is
a fuzzy variable defined by: ∀x ∈ R, (ξ + η)(x) = ξ(x) + η(x) and αξ is a
fuzzy variable defined by: ∀x ∈ R, (αξ)(x) = αξ(x).
A fuzzy number ξ is a fuzzy variable that satisfies: ∃a, b, c, d ∈ R with a ≤
b ≤ c ≤ d such that (i) µ is upper semi-continuous, (ii) ∀r ̸∈ [a, d], µ(r) = 0,
(iii) µ is increasing on [a, b] and decreasing on [c, d] and (iv) ∀r ∈ [b, c], µ(r) =
1. Thus, we denote it by ξ = (a, b, c, d). In the particular case where µ is
a straight line on [a, b] and [c, d], then ξ = (a, b, c, d) is the usual and well-
known trapezoidal fuzzy number. Then, we deduce  analytical expressions


x−a
if a ≤ x≤b
 b−a
1, if b ≤ x ≤ c
of the membership functions of ξ : ∀x ∈ R, µ(x) = .


x−d
if c ≤ x ≤ d
 c−d
0 elsewhere
If b = c, then ξ = (a, b, d) is a triangular fuzzy number.
Figure 1 displays a trapezoidal fuzzy variable and a triangular one.

3
2 Preliminaries

Figure 1: On right trapezoidal fuzzy variable (1, 2, 3, 4) and on left

triangular fuzzy variable (1, 3.5, 4).

In the following, we recall classical fuzzy measures proposed in the liter-

ature to deal with imprecision.
Let B ⊂ X and ξ : X → R be a fuzzy variable whose membership function
µ, that means, each element x of X is associated to a real number ξ(x)
by means of ξ with the membership grade µ(ξ(x)). We have the follow-
ing classical and well-known measures: the Possibility measure defined by
P os(B) = supx∈B µ(ξ(x)) (Zadeh [12]); the Necessity measure defined by
N ec(B) = 1 − P os(B c ) = inf x∈B µ(ξ(x)) (Zadeh [12]) and the Credibility
measure defined by Cred(B) = 21 [P os(B) + N ec(B)] (Liu [6]).
Yang and Iwamura [11] introduced a measure, denoted mλ and defined by
for λ ∈ [0, 1],

mλ (B) = λ sup µB (x)) + (1 − λ) inf µB (x)) (1)

x∈X x∈X

or equivalently, we have:

mλ ({ξ ∈ B}) = λP os({ξ ∈ B}) + (1 − λ)N ec({ξ ∈ B}).

In particular, we have: For λ = 0 (resp. λ = 1, resp. λ = 12 ) mλ = N ec

(resp. mλ = P os, resp. mλ = Cr). They established some axioms of mλ
such as universality, subadditivity for λ ≥ 21 and monotonicity. Notice that
if P(X) is the power set of X, then the triplet (X, P(X), mλ ) is called a
λ-fuzzy space.
In addition, we have the following two new axioms of mλ .

1.  
λmλ (B) + (1 − λ)mλ (B c ) = λ
 or  (2)
(1 − λ)mλ (B) + λmλ (B ) = λ
c

4
3 Characteristics of a fuzzy variable with respect to the mλ -measure

2. mλ (∪i∈I Bi ) = supi∈I mλ (Bi ) if supi∈I mλ (Bi ) ≤ λ where (Bi )i∈I is a

countable family of subsets of X and I ⊂ N.

When λ = 12 , the Axiom defined by (2) becomes the duality axiom of credi-
bility measure defined by Liu [6] as Cr(B) + Cr(B c ) = 1.

Let us end this Section by evaluating the measure of events ξ ≥ t and

ξ ≤ t for t ∈ R when ξ is a trapezoidal fuzzy variable.

Example 1. mλ {ξ ≥ t} = λ supr≥t µ(r) + (1 − λ)(1 − supr<t µ(r)) and

mλ {ξ ≤ t} = λ supr≤t µ(r) + (1 − λ)(1 − supr>t µ(r)).
We obtain:


 0 if d ≤ t,

 λ(d − t)

 if c ≤ t ≤ d,

 d−c
mλ {ξ ≥ t} = λ if b ≤ t < c,

 λ(t − a) + b − t



 if a ≤ t < b,

 b−a
1 if t < a


 1 if d ≤ t,

 λ(d − t) + t − c

 if c ≤ t ≤ d,

 d−c
mλ {ξ ≤ t} = λ if b ≤ t < c,

 λ(t − a)



 if a ≤ t < b,

 b−a
0 if t < a.

Throughout this paper, ξ is a normal fuzzy variable.

In the following Section, we introduce four main characteristics of a fuzzy

variable by means of mλ − measure. We compute them for trapezoidal and
triangular fuzzy variables and determine some of their properties.

3 Characteristics of a fuzzy variable with respect

to the mλ -measure
In the following Subsection, we introduce and study the expected value and
the variance of a fuzzy variable with respect to mλ − measure. Results of
our recent conference paper (Dzuché et al.[2]) are a part of the findings of
the Subsection.

5
3.1 Expected value and Variance

3.1 Expected value and Variance

Definition 1. Let ξ be a fuzzy variable and λ ∈ [0, 1]. The expected value
of ξ is defined by:
∫ 0 ∫ +∞
Eλ [ξ] = [mλ {ξ ≥ r} − 1]dr + mλ {ξ ≥ r}dr (3)
−∞ 0

with the condition that, at least one of the two integrals is finite.

Remark 1. 1) When λ = 21 , we obtain the expected value defined by Liu [6]

by means of the credibility measure as follows:
∫ +∞ ∫ 0
E 1 [ξ] = Cr{ξ ≥ r}dr − Cr{ξ ≤ r}dr
2
0 −∞

provided that, at least one of the two integrals is finite.

2) Liu [6] introduced for a normalized fuzzy variable, the upper expected
value of ξ and the lower expected value of ξ respectively denoted E[ξ] and
E[ξ] and defined by:
∫ +∞ ∫ 0
E[ξ] = P os{ξ ≥ r}dr − N ec{ξ ≤ r}]dr (4)
0 −∞

and ∫ ∫
+∞ 0
E[ξ] = N ec{ξ ≥ r}dr − P os{ξ ≤ r}]dr. (5)
0 −∞
We have:

Eλ [ξ] = λE[ξ] + (1 − λ)E[ξ] (6)

We now compute the expected values of trapezoidal and triangular fuzzy

variables.

Example 2. - The expected value of a trapezoidal fuzzy variable ξ = (a, b, c, d)

based on the mλ − measure is defined by:
a+b c+d
Eλ [ξ] = (1 − λ) +λ .
2 2
- We deduce that the expected value of a triangular fuzzy variable ξ = (a, b, c)
is given by:
a c b
Eλ [ξ] = (1 − λ) + λ + .
2 2 2
1
- If λ = , we obtain expected values of Liu [6] respectively for trapezoidal
2
and triangular fuzzy variables, that is, E 1 [ξ] = a+b+c+d
4 and E 1 [ξ] = a+2b+c
4 .
2 2

6
3.1 Expected value and Variance

The following result establishes that the expected value of a trapezoidal

fuzzy variable is increasing with respect to λ.
Proposition 1. Let ξ = (a, b, c, d) be a trapezoidal fuzzy variable, λ1 , λ2 ∈
[0, 1].
If λ1 ≤ λ2 then Eλ1 [ξ] ≤ Eλ2 [ξ].
Proof : Let ξ = (a, b, c, d) be a trapezoidal fuzzy variable and let us set:
f (λ) = (1 − λ) a+b ′
2 + λ 2 . We have: f (λ) = 2 + 2 ≥ 0 since c ≥ a and
c+d c−a d−b

d ≥ b. So f = Eλ [ξ] is increasing with respect to λ. 

We end this Subsection by introducing the Variance of a fuzzy variable

and we compute it in some cases.
Definition 2. Let ξ be a fuzzy variable such that E[ξ] = eλ and λ ∈ [0, 1] .
The variance of ξ is defined by:
∫ +∞
Vλ [ξ] = E[(ξ − eλ ) ] =
2
mλ {(ξ − eλ )2 ≥ r}dr. (7)
0

In the following Example, we compute the variance of a trapezoidal fuzzy

variable and a triangular one.
Example 3. 1. ξ = (a, b, c, d) is a trapezoidal fuzzy variable with ex-
pected value eλ .
β(eλ − a) − α(d − eλ )
We set: α = b − a, γ = c − b, β = d − c, t = and
β−α
without loss of generality, we assume that: b < eλ < c, γ < α < β,
c − eλ < eλ − a < d − eλ and c − eλ < t < eλ − a. We obtain:
Vλ [ξ] = λ(c − eλ )2 + αλ [(eλ − a)(t2 − (c − eλ )2 )
− 32 (t3 − (c − eλ )3 )] + βλ [(d − eλ )((eλ − a)2 − t2 ) − 32 ((eλ − a)3 − t3 )]
+ βλ [ 13 (d − eλ )3 ) + 32 (eλ − a)3 − (eλ − a)2 (d − eλ )].

2. ξ = (a, b, c) is a triangular fuzzy variable with expected value eλ .

β(eλ − a) − α(c − eλ )
We set: α = b − a, β = c − b, t = and without
β−α
loss of generality, we assume that: b > eλ and α > β. We obtain:
Vλ [ξ] = (b − eλ )2 − 1−λ α [(b − eλ ) (eλ − a) + 3 (b − eλ ) ]
2 2 3

+ βλ [(c − eλ )(t2 − (b − eλ )2 ) − 23 (t3 − (b − eλ )3 )] + αλ [(eλ − a)((c − eλ )2 −

t2 ) − 23 ((c − eλ )3 − t3 )]
+ αλ [ 13 (eλ − a)3 + 23 (c − eλ )3 − (c − eλ )2 (eλ − a)].
The following result establishes useful properties (linearity and homo-
geneity) of the expected value and the variance respectively with respect to
the mλ − measure.
Proposition 2. Let ξ and η be two fuzzy variables such that Eλ [ξ] < ∞ and
Eλ [η] < ∞, µ a nonnegative real number, ν a real number and λ ∈ [0, 1].

7
3.2 Skewness and Kurtosis

{
Eλ [ξ + η] = Eλ [ξ] + Eλ [η]
1. .
Eλ [µξ] = µEλ [ξ]

2. Vλ [µξ + ν] = µ2 Vλ [ξ].
Proof: 1) The proof of the linearity is given in the Section Appendix.
2) Since αξ + β is a fuzzy variable; on the other hand, by Definition 2 and
the first Proposition 2 , we have:
Vλ [µξ + ν] = Eλ [(µξ + ν − Eλ [µξ + ν])2 ] = Eλ [(µξ + ν − µEλ [ξ] − ν)2 ]. Thus,
we have: Vλ [µξ + ν] = Eλ [(µξ − µEλ [ξ])2 ] = Eλ [µ2 (ξ − Eλ [ξ])2 ]. Finally, by
the first result of Proposition 2, we have: Vλ [µξ + ν] = µ2 Eλ [(ξ − Eλ [ξ])2 ] =
µ2 Vλ [ξ]. 

Throughout this paper, we introduce new concepts and establish new

results.

In the following Subsection, we introduce and study the Skewness and

Kurtosis of a fuzzy variable with respect to the mλ - measure.

3.2 Skewness and Kurtosis

Definition 3. Let ξ be a fuzzy variable such that Eλ [ξ] = eλ and λ ∈ [0, 1] .
1) The skewness of ξ is defined by:
∫ 0 ∫ +∞
3
SKλ [ξ] = Eλ [(ξ−eλ ) ] = [mλ {(ξ−eλ ) ≤ r}−1]dr+
3
mλ {(ξ−eλ )3 ≥ r}dr
−∞ 0
(8)
2) The kurtosis of ξ is defined by:
∫ +∞
Kλ [ξ] = Eλ [(ξ − eλ )4 ] = mλ {(ξ − eλ )4 ≥ r}dr. (9)
0

We now compute skewness and kurtosis of a trapezoidal fuzzy variable

and a triangular one with expected value eλ .
Example 4. 1. For a trapezoidal fuzzy variable ξ = (a, b, c, d).
The Skewness SKλ [ξ] of ξ is given by:
1 λ
SKλ [ξ] = [(1−λ)(b−eλ )4 −λ(a−eλ )4 ]+ [(d−eλ )4 −(c−eλ )4 ].
4(b − a) 4(d − c)

2. For a triangular fuzzy variable ξ = (a, b, c).

The Skewness SKλ [ξ] of ξ is given by:
1 λ
SKλ [ξ] = [(1−λ)(b−eλ )4 −λ(a−eλ )4 ]+ [(c−eλ )4 −(b−eλ )4 ].
4(b − a) 4(c − b)
1
If λ = , we obtain Li et al.’s result on Skewness ([5]).
2

8
3.3 Semi-variance and semi-kurtosis

In the following Example, we compute kurtosis of a trapezoidal fuzzy

variable and a triangular one.

Example 5. 1. ξ = (a, b, c, d) is a trapezoidal fuzzy variable with ex-

pected value eλ .
β(eλ − a) − α(d − eλ )
We set: α = b − a, γ = c − b, β = d − c, t = and
β−α
without loss of generality, we assume that: b < eλ < c, γ < α < β,
c − eλ < eλ − a < d − eλ and c − eλ < t < eλ − a. We obtain:
Kλ [ξ] = λ(c − eλ )4 + αλ [(eλ − a)(t4 − (c − eλ )4 )
− 54 (t5 − (c − eλ )5 )] + βλ [(d − eλ )((eλ − a)4 − t4 ) − 54 ((eλ − a)5 − t5 )]
+ βλ [ 15 (d − eλ )5 ) + 54 (eλ − a)5 − (eλ − a)4 (d − eλ )].

2. ξ = (a, b, c) is a triangular fuzzy variable with expected value eλ .

β(eλ − a) − α(c − eλ )
We set: α = b − a, β = c − b, t = and without
β−α
loss of generality, we assume that: b > eλ and α > β. We obtain:
Kλ [ξ] = (b − eλ )4 − 1−λ α [(b − eλ ) (eλ − a) + 5 (b − eλ ) ]
4 4 5

+ β [(c − eλ )(t − (b − eλ ) ) − 5 (t − (b − eλ ) )] + α [(eλ − a)((c − eλ )4 −

λ 4 4 4 5 5 λ

t4 ) − 45 ((c − eλ )5 − t5 )]
+ αλ [ 15 (eλ − a)5 + 45 (c − eλ )5 − (c − eλ )4 (eλ − a)].

We end this Subsection by establishing some of their properties.

Proposition 3. Let ξ be a fuzzy variable, µ and ν two real numbers such

that µ ≥ 0 and λ ∈ [0, 1]. We have:
1) Sλ [µξ + ν] = µ3 Sλ [ξ].
2) Kλ [µξ + ν] = µ4 Kλ [ξ].

Proof : The proof of this Proposition is similar to the proof of the sec-
ond result of Proposition 2. 

In the following Subsection, we introduce semi-variance and semi-kurtosis

of a fuzzy variable ξ such that Eλ [ξ] = eλ and λ ∈ [0, 1].

3.3 Semi-variance and semi-kurtosis

Definition 4. 1) The semi-variance of ξ is defined by:
∫ +∞
− 2
Vλ [ξ] = Eλ [[(ξ − eλ ) ] ] =
S
mλ {[(ξ − eλ )− ]2 ≥ r}dr. (10)
0

2) The semi-kurtosis of ξ is defined by:

∫ +∞
− 4
Kλ [ξ] = Eλ [[(ξ − eλ ) ] ] =
S
mλ {[(ξ − eλ )− ]4 ≥ r}dr. (11)
0

9
3.3 Semi-variance and semi-kurtosis

Example 6. 1. For a trapezoidal fuzzy variable ξ = (a, b, c, d), we obtain:

a)

(eλ − a)3 max[0; (eλ − b)3 ] max[0; (eλ − c)3 ]

VλS [ξ] = λ −λ + (1 − λ) .
3(b − a) 3(b − a) 3(d − c)
1
If λ = , we obtain Huang’s result on semi-variance [4].
2
b)

(eλ − a)5 max[0; (eλ − b)5 ] max[0; (eλ − c)5 ]

KλS [ξ] = λ −λ + (1 − λ) .
5(b − a) 5(b − a) 5(d − c)
1
If λ = , we obtain Sadefo et al.’s result on semi-kurtosis [9].
2
2. For a triangular fuzzy variable ξ = (a, b, c), we obtain:
a)

(eλ − a)3 max[0; (eλ − b)3 ] max[0; (eλ − b)3 ]

VλS [ξ] = λ −λ + (1 − λ) .
3(b − a) 3(b − a) 3(c − b)
1
If λ = , we obtain Huang’s result on semi-variance [4].
2
b)

(eλ − a)5 max[0; (eλ − b)5 ] max[0; (eλ − b)5 ]

KλS [ξ] = λ −λ + (1 − λ) .
5(b − a) 5(b − a) 5(c − b)
1
If λ = , we obtain Sadefo et al.’s result on semi-kurtosis [9].
2
The following result justifies that the risk’s level of a trapezoidal fuzzy
variable (defined by semi-variance or semi-kurtosis) depend on the optimistic
parameter and on the expected return position with respect to the support
of the variable. Its proof is given in the Appendix.

Proposition 4. Let ξ = (a, b, c, d) be a trapezoidal fuzzy variable with ex-

pected value eλ and λ ∈ [0, 1].

1. If a ≤ eλ ≤ b or b ≤ eλ ≤ c then VλS [ξ] and KλS [ξ] increases with

respect to λ.

2. If c ≤ eλ ≤ d then:
- VλS [ξ] increases with respect to λ when λ ≤ 34 .
- KλS [ξ] increases with respect to λ when λ ≤ 56 .

Remark 2.

10
3.4 Some properties of fuzzy variables characteristics

For the cases where a ≤ eλ ≤ b or b ≤ eλ ≤ c, the trapezoidal fuzzy variable

describes almost either lower expected returns or average expected returns.
That is the reason why the more the investor is confident (λ increases), the
more he is exposed to a high risk’s level (semi-variance and semi-kurtosis
are greater).
For c ≤ eλ ≤ d, the trapezoidal fuzzy variable describes almost either higher
expected returns. Thus, the risk remains increasing for an optimistic maxi-
mum’s level ( 34 for semi-variance and 56 for semi-kurtosis).
For λ > 43 (resp. λ > 56 ), the monotonicity of the semi-variance of the trape-
zoidal fuzzy variable (resp. its semi-kurtosis) depends on α, β, γ = c−b. That
means, the investor can feel himself secured with high values of returns. In
that, risk cannot remain increasing with respect to the optimistic factor, it
will necessarily depend on returns’ spread (α, β, γ).
In the following, we establish some properties or relationships of charac-
teristics of a fuzzy variable ξ such that Eλ [ξ] = eλ and λ ∈ [0, 1].

3.4 Some properties of fuzzy variables characteristics

The following result justifies that variance (respectively kurtosis) is greater
than semi-variance (respectively semi-kurtosis). In addition, it justifies that
they are equal for symmetric fuzzy variables. Its proof is given in the Ap-
pendix.
Proposition 5. 1) We have: 0 ≤ VλS [ξ] ≤ Vλ [ξ] and 0 ≤ KλS [ξ] ≤ Kλ [ξ].
2) Furthermore, if ξ is symmetric with respect to eλ , that is, ∀r ∈ R, µ(eλ −
r) = µ(eλ + r), then { S
Vλ [ξ] = Vλ [ξ]
.
KλS [ξ] = Kλ [ξ]
The following result establishes necessary and sufficient conditions under
which Variance and Kurtosis are null for some values of λ.
Proposition 6. If λ ∈ [ 21 , 1[, then
{
Vλ [ξ] = 0 ⇔ mλ {ξ = eλ } = 1
.
Kλ [ξ] = 0 ⇔ mλ {ξ = eλ } = 1
To prove this result, we need the following Lemma which is proved in
the Appendix.
Lemma 1. Let A ⊂ X be an  event and λ ∈ [0, 1].
 0 if λ = 0
1) mλ (A) = 0 ⇒ mλ (A ) =
c λ
or 1 if λ ∈]0, 12 ]
 1−λ
1 if λ ∈] 12 , 1]
2) Furthermore, for λ ∈ [ 12 , 1[, we have:
mλ (A) = 0 ⇔ mλ (Ac ) = 1.

11
4 Application for portfolio selection

We now justify the Proposition.

Proof of Proposition 6: Let ξ be a fuzzy variable with Eλ [ξ] = eλ and
λ ∈ [ 12 , 1[. Let us prove that Vλ [ξ] = 0 ⇔ mλ {ξ = eλ } = 1.
(⇐): Let us assume that mλ {ξ = eλ } = 1.
It is obvious that mλ {ξ − eλ = 0} = 1 if and only if mλ {(ξ − eλ )2 } = 1.
By Lemma 1, we obtain mλ {(ξ − eλ )2 ̸= 0} = 0.
Let r > 0, we have:
mλ {(ξ − eλ )2 > r} ≤ {mλ {(ξ − eλ )2 > 0} ≤ {mλ {(ξ − eλ )2 ̸= 0} = 0.
This implies that m∫λ {(ξ − eλ )2 > r} = 0, ∀r > 0.
+∞
Therefore, Vλ [ξ] = 0 mλ {(ξ − eλ )2 ≥ r}dr = 0.
(⇒): Let us assume that Vλ [ξ] = 0.
∫ +∞
Since mλ takes values in [0;1], 0 mλ {(ξ − eλ )2 ≥ r}dr = 0 implies that
mλ {(ξ − eλ )2 ≥ r} = 0, ∀r > 0, that is mλ {(ξ − eλ )2 ̸= 0} = 0.
By Lemma 1, we have: mλ {(ξ − eλ )2 = 0} = 1 and we deduce that mλ {ξ −
eλ = 0} = 1, that is, mλ {ξ = eλ } = 1.
In a similar way, it is easy to prove that Kλ [ξ] = 0 ⇔ mλ {ξ = eλ } = 1. 
In the following Section, we propose two models and implement them for
the determination of optimal portfolios in Finance.

4 Application for portfolio selection

4.1 Optimization models
Let us consider an investor who likes to invest his capital in n securities
∑nthe proportion x1 , x2 ,..., xn such that ∀i ∈ {1, 2, ..., n}, xi ∈ [0, 1] and
in
i=1 xi = 1. It is well-known that an investment of a part xi of the capital
in the ith security generates a return denoted by xi ξi which is not currently
known. Making up such investment consists on constituting a portfolio
((xi , ξi ))1≤i≤n where the n fuzzy variables x1 ξ1 , ..., xn ξn are future returns
of the n securities and the fuzzy variable ξ = ξ1 x1 + ξ2 x2 + ... + ξn xn is
the total future return or the portfolio future return. With respect to the
credibility measure, Huang [4], Li et al. [5], Sadefo et al. [9] assumed
that ξi is a fuzzy variable and they proposed models based on parameters
(mean, variance and semi-variance for Huang; mean, variance and skew-
ness for Li et al.; mean, variance, skewness and kurtosis for Sadefo et al.)
in order to determine optimal portfolios. In this Section, we propose two
models based on parameters with respect to mλ to solve the same ques-
tion of determination of best portfolios. Those new models give a family
of solutions which are optimal portfolios and generalize more recent models
proposed by Sadefo et al. [9] in the particular case where λ = 12 . For that,
we consider the first family of seven assets returns described by the fol-
lowing triangular fuzzy variables proposed by Huang [4] and used by many
authors: ξ1 = (−0.3, 1.8, 2.3), ξ2 = (−0.4, 2.0, 2.2), ξ3 = (−0.5, 1.9, 2.7),
ξ4 = (−0.6, 2.2, 2.8), ξ5 = (−0.7, 2.4, 2.7), ξ6 = (−0.8, 2.5, 3.0) , ξ7 =

12
4.2 Implementation of the two models with Huang’s data

(−0.6, 1.8, 3.0) and the second family of ten assets returns described by the
following trapezoidal fuzzy variables proposed by Hasuike et al. [3] (written
as (a, b, c, d)): ξ1 = (−0.362, −0.123, 0.005, 0.873), ξ2 = (−0.37, −0.069, 0.069, 0.536),
ξ3 = (−0.329, −0.129, 0.025, 0.738), ξ4 = (−0.193, 0.005, 0.177, 0.412), ξ5 =
(−0.299, −0.082, 0.114, 0.437), ξ6 = (−0.342, −0.052, 0.108, 0.49) ,
ξ7 = (−0.292, −0.056, 0.067, 0.436), ξ8 = (−0.25, 0.060, 0.193, 0.649), ξ9 =
(−0.405, −0.093, 0.130, 0.756) , ξ10 = (−0.554, 0.009, 0.236, 0.5). In addi-
tion, we set for each family of assets, three targets values t1 , t2 , t3 considered
respectively as the minimum benefit (expected value), the maximum risk
(variance) and the minimum skewness that the investor can bear. The two
proposed selection models for best portfolios of those assets are:


 minimize Kλ [xi1 ξi1 + xi2 ξi2 + ... + xik ξik ]



 subject to


 λ i1 i1 + xi2 ξi2 + ... + xik ξik ] ≥ t1
E [x ξ
Vλ [xi1 ξi1 + xi2 ξi2 + ... + xik ξik ] ≤ t2 (12)



 S [x ξ
λ i1 i1 + x ξ
i2 i2 + ... + x ξ
ik ik ] ≥ t 3


 xi1 + xi2 + ... + xik = 1


xik ≥ 0, k ∈ {7; 10}

and 

 minimize KλS [xi1 ξi1 + xi2 ξi2 + ... + xik ξik ]



 subject to


 Eλ [xi1 ξi1 + xi2 ξi2 + ... + xik ξik ] ≥ t1
Vλ [xi1 ξi1 + xi2 ξi2 + ... + xik ξik ] ≤ t2 (13)


 Sλ [xi1 ξi1 + xi2 ξi2 + ... + xik ξik ] ≥ t3




 x + xi2 + ... + xik = 1
 i1
xik ≥ 0, k ∈ {7; 10}
In the following paragraph, we implement the two models by using Huang’s
data.

4.2 Implementation of the two models with Huang’s data

In the following, we implement with Huang’s data the two previous models
for λ = 13 , λ = 12 and λ = 23 where the selected target values t1 , t2 , t3 are
given in Table 3.
For that, we set: ∀i ∈ {1, ..., 7}, ξi = (ai , bi , ci ) and the combination of those
seven triangular
∑7 fuzzy variables is also
∑7a triangular ∑fuzzy variable ∑
denoted
7 7
by ξ = i=1 ξi = (a, b, c) where a ∑ = i=1 ai , b = ∑ i=1 bi and c = i=1 ci ,
αi = bi − ai , βi = ci − bi and α = 7i=1 αi , β = 7i=1 βi .
We recall that Sadefo et al. [9] defined these characteristics in the particular
case where λ = 12 .
By implementing those formulas in Matlab through the previous pro-
posed models, we obtain (i) Table 1 which presents in each line (from the
second line) how the best portfolio is made up of in term of percentage of

13
4.2 Implementation of the two models with Huang’s data

the seven securities for a given value of λ and for either K, either K S as
objective function and, (ii) Table 2 gives characteristics (parameters) of best
portfolios of Table 1.

Asset i 1 (%) 2 (%) 3 (%) 4 (%) 5 (%) 6 (%) 7 (%)

1
λ = ( K) 0.00 63.39 0.00 0.00 36.61 0.00 0.00
3
1
λ = ( KS) 0.05 0.6471 0.00 0.05 31.33 3.85 0.00
3
1
λ = ( K) 20.04 0.00 0.00 79.89 0.00 0.07 0.00
2
1
λ = ( K S ) 20.00 0.00 0.00 80.00 0.00 0.00 0.00
2
2
λ = ( K) 42.50 0.00 0.00 57.50 0.00 0.00 0.00
3
2
λ = (K S ) 42.50 0.00 0.00 57.50 0.00 0.00 0.00
3
Table 1: Optimal selection from each model.

Mean Variance SV Skewness Kurtosis SK

λ = 13 ( K) 1.3005 0.4571 0.2482 -0.0014 0.6593 0.4880
λ = 13 ( K S ) 1.3004 0.4570 0.2482 −1.5336 × 10−5 0.6594 0.4875
λ = 12 ( K) 1.60 0.7018 0.6140 -0.6823 1.7290 1.6873
λ = 12 ( K S ) 1.60 0.7019 0.6141 -0.6823 1.7291 1.6872
λ = 23 ( K) 1.80 1.0510 1.0404 -1.6573 3.2213 3.2202
λ = 23 (K S ) 1.80 1.0510 1.0404 -1.6573 3.2213 3.2202

Table 2 : Comparison of the characteristics of different optimal portfolios.

Target values Mean:t1 Variance: t2 Skewness: t3

1
λ= 1.3 0.5 0
3
1
λ= 1.6 0.8 -0.6823
2
2
λ= 1.8 1.2 -2
3
Table 3: Selected target values depending on λ.

Table 1 gives strategies to invest in the seven assets in order to obtain

best portfolios for each of the three values of the parameter. More precisely,
no investment in ξ3 and ξ7 , greater investment in ξ1 and ξ4 for greater values
of λ.
We can observe the impact of the parameter λ variation through the follow-
ing histogram.

14
4.2 Implementation of the two models with Huang’s data

Figure 2: Illustration of fuzzy variables characteristics from different

models with Huang’s data.

According to the results of Table 2, we can make the following observa-

tions:
- The mean increases with respect to λ, it can be interpreted as follows:
the more the optimistic parameter is greater (the more self confident the
investor is), the greater is the expected return which represents its benefit.
In addition, risk measures such as variance, kurtosis, semi-variance, semi-
kurtosis increase with respect to λ. It means that, when the investor is self
confident in investment and he is looking for greater benefits, meanwhile he
is exposed to greater risk.
- The skewness is decreasing with respect to λ. It means that returns’ spread
is more greater on left of the mean when λ increases. Thus, the investor is
exposed to loss (negative returns) when he is more self confident.
- Notice that, for a given value of λ, it is not always possible to get invest-
ment proportions with some targets values. By varying λ as in Table 3, one
can obtain models that converge to a unique solution with some given target
values.
- For λ = 23 , the two proposed models coincide. For further research, it is
interesting to determine different optimistic parameter’s values or the min-
imum optimistic parameter’s value such that the two models coincide.

15
4.3 Implementation of the two models with Tokyo stock exchange data

To end this Section, we implement the two proposed models by using

real data as Tokyo stock exchange data.

4.3 Implementation of the two models with Tokyo stock

exchange data
We implement Tokyo stock exchange data through the two previous models
for λ = 14 , λ = 13 and λ = 12 . The target values t1 , t2 , t3 are given in Table 6.
Different optimal investment proportions and optimal characteristics are re-
spectively given in Tables 4 and 5.

Asset i 1 (%) 2 (%) 3 (%) 4 (%) 5 (%) 6 (%) 7 (%) 8 (%) 9 (%) 10 (%)
(Kλ= 1 ) 0.00 0.00 0.00 45.30 0.00 0.00 0.00 54.69 0.00 0.00
4
S
(Kλ= 1) 0.05 0.04 0.05 43.77 0.06 0.05 0.07 55.84 0.04 0.03
4
(Kλ= 1 ) 0.00 0.00 0.00 16.83 0.00 0.00 0.00 83.16 0.00 0.00
3
S
(Kλ= 1) 0.00 0.00 0.00 16.83 0.00 0.00 0.00 83.16 0.00 0.00
3
(Kλ= 1 ) 48.51 0.00 0.00 28.78 0.00 0.00 18.87 3.75 0.00 0.00
2
S
(Kλ= 1) 48.37 0.00 0.00 31.71 0.00 0.00 17.56 2.25 0.00 0.00
2

Table 4: Optimal selection from each model.

Mean Variance SV Skewness Kurtosis SK

λ = 14 ( K) 0.02 0.0262 0.0047 0.012 0.0049 1.6745 × 10−4
λ = 14 ( K S ) 0.0201 0.0264 0.0047 0.0122 0.0050 1.7094 × 10−4
λ = 13 ( K) 0.07 0.0285 0.0114 0.0141 0.0058 6.6001 × 10−4
λ = 13 ( K S ) 0.07 0.0286 0.0114 0.0142 0.0058 6.6330 × 10−4
λ = 12 ( K) 0.09 0.04 0.0389 0.01 0.0083 0.0037
λ = 12 (K S ) 0.09 0.04 0.0385 0.01 0.0081 0.0036

Table 5 : Comparison of the characteristics of different optimal portfolios.

Target values Mean:t1 Variance: t2 Skewness: t3

1
λ= 0.02 0.04 0.01
4
1
λ= 0.07 0.04 0.01
3
1
λ= 0.09 0.04 0.01
2
Table 6: Selected target values depending on λ.

As we can observe from Tables 5 and 2, optimal portfolios characteristics

except the skewness (skewness neither describes benefit nor risk), vary in the
same manner with respect to the parameter λ whatever data are described.

16
5 Concluding Remarks

The following histogram provides for Tokyo stock exchange data, clear ob-
servations of those characteristics variations according to the model and the
parameter λ.

Figure 3: Illustration of fuzzy variables characteristics from different

models with Tokyo stock exchange data.

5 Concluding Remarks
In this paper, we have determined two useful axioms of the mλ − measure of
fuzzy events combining possibility and necessity measures with an optimistic
parameter which describes investor attitude to make a decision in uncertain
situations. By means of the mλ -measure and its axioms, we have defined and
determined properties of some characteristics (four first moments and two
first semi-moments) of a fuzzy variable. These moments and semi-moments
have been determined and analyzed for trapezoidal and triangular fuzzy
variables. The obtained results generalize those obtained earlier by Liu [6],
Li et al. [5], Huang et al. [4] and Sadefo et al. [9]. Those theoretical results
have been illustrated in portfolio selection with fuzzy returns in Finance.
We have proposed and implemented two models which led to a huge variety
of solutions in portfolio selection according to some given target values. We
have displayed the impact of the variation of the optimistic parameter on
the investment’s risk and the benefit: the benefit and the risk increase with
respect to the optimistic parameter.
For further research, we propose to tackle some interesting questions among

17
6 Appendix

which: (i) the determination of k-moments and 2k-moments (for k ∈ N) of

a fuzzy variable (ii) the study of dominance on fuzzy variables based on the
new measure and (iii) the study of partial lower moments of a fuzzy variable
based on the new measure.

6 Appendix
Proof of Proposition 2 (linearity of the mean): According to the ex-
pressions recalled in relations (4) and (5) and the notions of optimistic and
pessimistic functions respectively given by:
Xsup (α) = sup{r/P os{x ∈ X/ξ(x) ≥ r} ≥ α} and Xsup (α) = sup{r/P os{x ∈
X/ξ(x) ≥ r} ≥ α} with α ∈ (0, 1], Liu and Liu [7] proved that:
∫1 ∫1
1. E[ξ] = 0 Xsup (α)dα and E[ξ] = 0 Xinf (α)dα.

2. E[ξ + η] = E[ξ] + E[η] and E[ξ + η] = E[ξ] + E[η].

3. For a ≥ 0, E[aξ] = aE[ξ] and E[aξ] = aE[ξ].

4. For a ≤ 0, E[aξ] = aE[ξ] and E[aξ] = aE[ξ].

Furthermore, according to relation (6), we have:

For λ ∈ R, Eλ [ξ] = λE[ξ] + (1 − λ)E[ξ] and thus:
Eλ [ξ + η] = λE[ξ + η] + (1 − λ)E[ξ + η], that is,
Eλ [ξ + η] = [λE[ξ] + (1 − λ)E[ξ]] + [λE[η] + (1 − λ)E[η]] = Eλ [ξ] + Eλ [η].
On the other hand, we have for a nonnegative real number α:
Eλ [αξ] = λE[αξ] + (1 − λ)E[αξ]=α(λE[ξ] + (1 − λ)E[ξ]]) = αEλ [ξ].

Proof of Proposition 4: Let ξ = (a, b, c, d) be a trapezoidal fuzzy

variable with expected value eλ and λ ∈ [0, 1].
−a)
and KλS [ξ] = λ (eλ −a)
3 5
• If a ≤ eλ ≤ b then VλS [ξ] = λ (eλ3α 5 .
(eλ −a)3 e′λ (eλ −a)2 (eλ −a)5
We have: (V S )′λ [ξ] = 3α +λ α ≥ 0 and (K S )′λ [ξ] = 5α +
e′λ (eλ −a)4
λ α ≥ 0.
−a) −b)
and KλS [ξ] = λ (eλ −a)
3 3 5
• If b ≤ eλ ≤ c then VλS [ξ] = λ (eλ3α − λ (eλ3α 5 −
−b) 5
λ (eλ5α .
(eλ −a)3 −b)3 e′ (e −a)2 e′ (e −b)2
We have: (V S )′λ [ξ] = 3α − (eλ3α + λ λ λα − λ λ λα ≥0
(eλ −a)5 −a)5 e′ (e −a)4 e′ (e −b)4
and (K S )′λ [ξ] = 5α − (eλ5α + λ λ λα − λ λ λα ≥ 0.
−a) −b) 3 −c) 3 3
• If c ≤ eλ ≤ d then VλS [ξ] = λ (eλ3α − λ (eλ3α + (1 − λ) (eλ3β and
KλS [ξ] = λ (eλ −a) −b) −c)
5 5 5
5 − λ (eλ5α + (1 − λ) (eλ5β .
(eλ −a)3 (eλ −b)3 e′λ (eλ −a)2 e′λ (eλ −b)2
We have: (V S )′λ [ξ] = 3α − 3α +λ α −λ α −

18
6 Appendix

(eλ −c)3 e′ (e −c)2 −a)5 (eλ −a)5 e′ (e −a)4

3β +(1−λ) λ λβ and (K S )′λ [ξ] = (eλ5α − 5α +λ λ λα −
e′λ (eλ −b)4 (eλ −c)5 e′λ (eλ −c)4
λ α − 5α + (1 − λ) β .
We easily check that:
λ ≤ 34 ⇒ (V S )′λ [ξ] ≥ 0 and λ ≤ 5
6 ⇒ (K S )′λ [ξ] ≥ 0. 
Proof of Proposition 5 1) Let us prove that 0 ≤ VλS [ξ] ≤ Vλ [ξ]. Let
x ∈ X and r ∈{R. We have:
(ξ − eλ )2 if ξ ≤ eλ
[(ξ − eλ )− ]2 = . This definition leads us to two cases:
0 if ξ > eλ
i) First case: ξ(x) ≤ eλ .
We have: [(ξ(x) − eλ )− ]2 = (ξ(x) − eλ )2 .
Therefore, [(ξ(x) − eλ )− ]2 ≥ r ⇔ (ξ(x) − eλ )2 ≥ r.
ii) Second case: ξ(x) > eλ .
We have: [(ξ(x)−eλ )− ]2 = 0 which implies that (ξ(x)−eλ )2 ≥ [(ξ(x)−eλ )− ]2 .
Thus, [(ξ(x) − eλ )− ]2 ≥ r implies (ξ(x) − eλ )2 ≥ r.
It follows that: ∀(x, r) ∈ X ×R, {x/[(ξ(x)−eλ )− ]2 ≥ r} ⊆ {x/(ξ(x)−eλ )2 ≥
r}.
By the fact that mλ is monotone, we obtain:
∀r ∈ R, mλ {[(ξ(x) − eλ )− ]2 ≥ r} ≤ mλ {(ξ(x)∫− eλ )2 ≥ r}.
∫ +∞
Finally, Vλ [ξ] = 0 mλ {(ξ − eλ )2 ≥ r}dr ≥ 0 mλ {[(ξ − eλ )− ]2 ≥ r}dr =
+∞
S
Vλ [ξ].
In the same manner, one can prove that 0 ≤ kλS [ξ] ≤ Kλ [ξ].

2) Let ξ be a symmetric fuzzy variable. Let us prove that VλS [ξ] = Vλ [ξ].
∫ +∞
Let us write: VλS [ξ] = Eλ [[(ξ − eλ )− ]2 ] = 0 ⇕λ {[(ξ − eλ )− ]2 ≥ r}dr =
∫ +∞ √ √
⇕λ {(ξ − eλ )− ∈] − ∞; − r] ∪ [ r; +∞[}dr and
0 ∫ +∞ ∫ +∞
Vλ [ξ] = Eλ [(ξ − eλ )2 ] = 0 ⇕λ {(ξ − eλ )2 ≥ r}dr = 0 ⇕λ {(ξ − eλ ) ∈
√ √
] − ∞; − r] ∪ [ r; +∞[}dr.
If µ and ν are respectively membership { functions of fuzzy variables ξ − eλ
µ if ξ − eλ ≤ 0
and (ξ − eλ )− , then we have ν = . Let us observe that
0 otherwise
ξ is symmetric with respect to eλ if and only if ξ − eλ is symmetric with
respect to 0. We have, ∀r > 0:
√ √
Mλ {(ξ−eλ ) ∈]−∞; − r]∪[ r; +∞[} = λ supx∈]−∞;−√r]∪[√r;+∞[ µ(x)+(1−
λ)(1−supx∈]−√r;−√r[ µ(x)) = λ max(supx∈]−∞;−√r] µ(x), supx∈[√r;+∞[ µ(x))+
(1 − λ)[1 − max(supx∈]−√r;0] µ(x), supx∈[0;√r[ µ(x))].
Let us notice that by fact that ξ is symmetric, we have: supx∈]−∞;−√r] µ(x) =
supx∈[√r;+∞[ µ(x) and supx∈]−√r;0] µ(x) = supx∈[0;√r[ µ(x). so, we have:
√ √
⇕λ {(ξ − eλ ) ∈] − ∞; − r] ∪ [ r; +∞[} = λ supx∈]−∞;−√r] µ(x) + (1 − λ)[1 −
supx∈]−√r;0] µ(x)).
However, supx∈]−∞;−√r] µ(x) = supx∈]−∞;−√r] ν(x) and supx∈]−√r;0] µ(x) =
supx∈]−√r;0] ν(x), that is:
√ √
⇕λ {(ξ − eλ ) ∈] − ∞; − r] ∪ [ r; +∞[} = λ supx∈]−∞;−√r] ν(x) + (1 − λ)[1 −
supx∈]−√r;0] ν(x)).

19
6 Appendix

√ √
By the fact that ν is null on the intervals [ r; +∞[ and [0; r[, we finally
have:
√ √
⇕λ {(ξ −eλ ) ∈]−∞; − r]∪[ r; +∞[} = λ supx∈]−∞;−√r]∪[√r;+∞[ ν(x)+(1−
√ √
λ)[1 − supx∈]−√r;−√r[ ν(x)) = ⇕λ {(ξ − eλ )− ∈] − ∞; − r] ∪ [ r; +∞[}.
Therefore, we get VλS [ξ] = Vλ [ξ].
By a similar way, one can prove that KλS [ξ] = Kλ [ξ]. 

Proof of Lemma 1: 1) Let us consider A ⊂ X such that Mλ (A) =

0 and λ ∈ [0, 1]. Let us recall the third axiom of the measure Mλ (A):
 λmλ (A) + (1 − λ)mλ (Ac ) = λ
or

(1 − λ)mλ (A) + λmλ (Ac ) = λ
• If
 λ = 0, cthen we obtain by Axiom 3:
 mλ (A ) = 0
or . That is, λmλ (Ac ) = 0.

0=0

• If λ ∈]0, 12 ], then we obtain by axiom 4:

 
 (1 − λ)mλ (Ac ) = λ λ
 mλ (Ac ) = 1−λ
or which implies that or .
 c  c
λmλ (A ) = λ mλ (A ) = 1
Furthermore, λ ∈]0, 21 ] ⇒ 0 < mλ (Ac ) = 1−λ
λ
≤ 1.

• If λ ∈] 12 , 1[, then we obtain by Axiom 4:

 
 (1 − λ)mλ (Ac ) = λ  mλ (Ac ) = 1−λλ

or which implies that or .

 c  c
λmλ (A ) = λ mλ (A ) = 1
However, λ ∈] 2 , 1[⇒ mλ (A ) = 1−λ > 1, which is wrong. So, we have
1 c λ

mλ (Ac ) = 1.
Finally,
 for λ = 1, we obtain by Axiom 3:
 0=1
or . However, 0 = 1 is wrong. So, λmλ (Ac ) = 1.
 c
mλ (A ) = 1

2) Let us consider A ⊂ X and λ ∈ [ 12 , 1[.

⇒) If mλ (A) = 0, then according to Lemma 1, we have mλ (Ac ) = 1.
⇐) If mλ (Ac ) = 1, then by Axiom 3, we have: 

 λmλ (A) + (1 − λ) = λ  mλ (A) = 2λ−1
λ
or which implies that or .
 
(1 − λ)mλ (A) + λ = λ mλ (A) = 0
However, λ ∈ [ 21 , 1[⇒ mλ (A) = 2λ−1
λ < 0, which is wrong. So, we have
mλ (A) = 0. Hence the result. 

20
REFERENCES REFERENCES

References
[1] Dai C., Cai Y.P., Li Y.P, Sun W., Wang X.W., and Guo H.C. (2014) Op-
timal strategies for carbon capture, utilization and storage based on an
inexact mλ − measure fuzzy chance-constrained programming. Energy,
78, 465-478.
[2] Dzuché J., Fono L.A., Tassak D.C. and Sadefo K.J. (2017) Expected
value and variance of a fuzzy variable based on a new fuzzy measure.
International Statistical Institute Proceeding, the 61st World Statistical
Congress, Marrakeck, Morocco, 15-17 July 2017.
[3] Hasuike T., Katagiri H. and Ishii H. (2009) Portfolio selection problems
with random fuzzy variable return. Fuzzy Sets and Systems, 2579-2596.
[4] Huang X. (2008) Mean-semivariance models for fuzzy portfolio selec-
tion.Journal of Computational and Applied Mathematics, 217, 1-8.
[5] Li X., Qin Z. and Kar S. (2010) Mean-variance-skewness model for port-
folio selection with fuzzy returns. European Journal of Operational Re-
search, 202, 239-247.
[6] Liu B. and Liu Y.K. (2002) Expected value of fuzzy variable and fuzzy
expected value models: IEEE Transactions on Fuzzy Systems, 10, 445-
450.
[7] Liu B. and Liu Y.K. (2003) Expected value operator of random fuzzy
variable and random fuzzy expected models, International Journal of
Uncertainty, Fuzziness and Knowledge-Based Systems, 11, 195-215
[8] Peng J., Mok H. and Tse W. (2005) Fuzzy Dominance Based on Credi-
bility Distributions. Springer-Verlag, Berlin, 295-303.
[9] Sadefo K.J., Tassak D.C. and Fono L.A. (2012) Moments and semi-
moments for fuzzy portfolio selection. Insurance: Mathematics and Eco-
nomics 51, 517-530.
[10] Tassak D.C., Sadefo K.J., Fono L.A. and Andjiga N.G. (2017) Char-
acterization of order dominances on fuzzy variables for portfolio selec-
tion with fuzzy returns.Journal of Operational Research Society, doi
10.1057/s41274-016-0164-5.
[11] Yang L. and Iwamura K. (2008) Fuzzy Chance-Constrained Program-
ming with Linear Combination of Possibility Measure and Necessity mea-
sure. Applied Mathematical Sciences, 2 (46) 2271-2288.
[12] Zadeh L.A. (1978) Fuzzy set as a basis of theory of possibility. Fuzzy
Sets and Systems 1, 3-28.

21