You are on page 1of 6

Chaos, Solitons and Fractals 146 (2021) 110842

Contents lists available at ScienceDirect

Chaos, Solitons and Fractals


Nonlinear Science, and Nonequilibrium and Complex Phenomena
journal homepage: www.elsevier.com/locate/chaos

A new mean-variance-entropy model for uncertain portfolio


optimization with liquidity and diversification
Bo Li∗, Ranran Zhang
School of Applied Mathematics, Nanjing University of Finance and Economics, Nanjing 210023, China

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

Article history: This paper deals with a portfolio optimization problem with uncertain returns. Here, the returns of risky
Received 14 December 2020 assets are regarded as uncertain variables which are estimated by experienced experts. First, a mean-
Revised 9 February 2021
variance-entropy model for uncertain portfolio optimization problem is presented by taking into account
Accepted 1 March 2021
four criteria viz., return, risk, liquidity and diversification degree of portfolio. In our model, the invest-
Available online 11 March 2021
ment return is quantified by uncertain expected value, the investment risk is characterized by uncertain
Keywords: variance and entropy is used to measure the diversification degree of portfolio. Moreover, different from
Uncertain variable the previous bi-objective optimization model, our model achieves both the maximum return and the min-
Portfolio optimization imum risk in a single objective form by introducing a risk aversion factor and the dimensional influence
Return rate caused by different units is eliminated by normalization method. Then, two auxiliary portfolio selection
Diversification models are transformed into different equivalent deterministic models. Finally, a numerical simulation is
Liquidity
given to verify the effectiveness and practicality of our model.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction torical data about assets returns. However, it is sometimes hard


to achieve for complex financial markets. Based on fuzzy set the-
The purpose of portfolio selection is to choose an optimal port- ory [14], many researches have employed fuzzy variables to de-
folio according to certain standards. Markowitz [1] acted as pi- scribe experts’ subjective estimates, such as [15–20]. Among these,
oneer in proposing the famous mean-variance model in 1952, entropy and liquidity are widely used in fuzzy portfolio optimiza-
where variance was used as a risk measure. Since then, numer- tion models [21–25]. For instance, Mehlawat [22] constructed two
ous studies have been carried out based on mean-variance model, credibilistic mean-entropy models for portfolio optimization prob-
and a large number of them engaged in the extensions of mean- lem with multi-choice aspiration levels. Fang et al. [24] and Parra
variance model, such as [2–8]. In addition, entropy was also ap- et al. [25] both presented a portfolio model with considering three
plied to describe investment risk. Aksarayli and Pala [9] proposed a criteria: return, risk and liquidity.
new approach for polynomial goal programming which was based Nevertheless, when we use fuzzy variables to describe the re-
on a mean-variance-skewness kurtosis-entropy model. Bera and turn of risky assets, there will be a paradox (see Huang and
Park [10] introduced a cross-entropy measure as the objective Ying [26]). In order to better describe the subjective imprecise
function. Kapur and Kesavan [11] presented some applications of quantity, Liu [27] developed an uncertainty theory in 2007. At
the various entropy optimization principles. As we all know, en- present, numerous researchers have been working on the portfo-
tropy is not only a risk measurement tool, but also a tool for mea- lio optimization problems based on uncertainty theory [28–37]. As
suring the diversification degree of portfolio. Jana et al. [12] and a matter of fact, entropy has been also used in uncertain portfolio
Kapur [13] both introduced the entropy into their models for mea- optimization problems. Bhattacharyya [38] proposed an uncertain
suring the diversification degree of portfolio. multiple objective mean-entropy-skewness model. Gao [39] put
All the above literature assumes that the return of risky asset is forward a risk-free protection index model with an entropy con-
a random variable and then the probability theory is used to solve straint under uncertainty framework. Moreover, many researchers
the corresponding portfolio optimization problem. The premise of regarded liquidity as a constraint for uncertain portfolio models. In
using probability theory is that there is a large number of his- Zhai [40], turnover rate was used to describe liquidity and it was
regarded as an uncertain variable. However, there are few studies
related to both diversification and liquidity constraints.

Corresponding author. In this paper, diversification and liquidity will be considered as
E-mail address: libnust@163.com (B. Li). constraints in the uncertain portfolio optimization problem. The

https://doi.org/10.1016/j.chaos.2021.110842
0960-0779/© 2021 Elsevier Ltd. All rights reserved.
B. Li and R. Zhang Chaos, Solitons and Fractals 146 (2021) 110842

major contributions of this paper are: Firstly, diversification and Theorem 4 (Liu [27]). Assume ξ1 , ξ2 , . . . , ξn are independent
liquidity are both studied in uncertain environment. Secondly, we uncertain variables with regular uncertainty distributions 1 ,
introduce a risk aversion factor to transform the bi-objective opti- 2 , . . . , n , respectively. If f (ξ1 , ξ2 , . . . ξn ) is strictly increasing with
mization model into a single objective optimization model and the respect to ξ1 , ξ2 , . . . , ξn and strictly decreasing with respect to
dimensional influence caused by different units is eliminated by ξm+1 , ξm+2 , . . . , ξn , then ξ = f (ξ1 , ξ2 , . . . , ξn )
normalization method. The rest of this paper is organized as fol-
1
lows. In Section 2, we review some basic concepts in uncertainty E[ξ ] = 1 (α ), . . . , m (α ), m+1 (1 − α ), . . . , n (1 − α ))d α .
f (−1 −1 −1 −1
theory. A mean-variance-entropy model for uncertain portfolio op- 0
timization problem is established in Section 3. In Section 4, two
Theorem 5 (Liu [27]). Let ξ be an uncertain variable with regular
auxiliary portfolio selection models are transformed into different
uncertainty distribution , then
equivalent deterministic models. In Section 5, a numerical simula-
tion is carried out to show the effectiveness and practicality of our 1

model. Finally, some conclusions are given. E[ξ ] = −1 (α )dα .


0

Remark 1. An uncertain variable is essentially a measurable func-


2. Preliminary
tion from an uncertainty space to the set of real numbers and a
random variable is a measurable function from a probability space
Let  be a nonempty set, and L a σ -algebra over  . Each el-
to the set of real numbers. Moreover, they have different theoreti-
ement  ∈ L is called an event. A set function M defined on the
cal bases and obey different operational laws.
σ -algebra over L is called an uncertain measure if it satisfies the
following axioms: (normality axiom) M{ } = 1 for the universal
3. Uncertain mean-variance-entropy model
set  ; (duality axiom) M{} + M{c } = 1 for any event ; (sub-
 ∞
additivity axiom) M{ ∞ i=1 i } ≤ i=1 M{i } for every countable
Here, we consider an uncertain portfolio optimization problem
sequence of events 1 , 2 , . . ..
with n risky assets. Let xi , i = 1, 2, . . . , n, be the investment pro-
The triplet ( , L, M ) is called an uncertainty space. A prod-
portion in risky asset i, and ξi , i = 1, 2, . . . , n, be the uncertain re-
uct uncertain measure M is defined by Liu [41] to produce
turn rate of risky asset i. Assume that uncertain variables ξi and ξ j
an uncertain measure of compound event: (product axiom) Let
(i, j = 1, 2, . . . , n and i = j) are independent and the short selling
(k , Lk , Mk ) be uncertainty spaces for k = 1, 2, . . .. Then the prod-
of risky assets is not allowed, i.e., xi ≥ 0, i = 1, 2, . . . , n.
uct uncertain measure M is an uncertain measure satisfying
  The expected return of portfolio x = (x1 , x2 , x3 , . . . , xn ) is given

 ∞
 by
M k = Mk {k },
E[ξ1 x1 + ξ2 x2 + ξ3 x3 + . . . + ξn xn ]. (3.1)
k=1 k=1

where k are arbitrarily chosen events from Lk for k = 1, 2, . . . , As we know, variance is widely used as a risk measure. Here, the
respectively. variance of portfolio x can be written as:
An uncertain variable is a function ξ from an uncertainty space V [ ξ1 x 1 + ξ2 x 2 + ξ3 x 3 + . . . + ξ n x n ] . (3.2)
( , L, M ) to the set of real numbers such that for any Borel set of
real numbers, the set {ξ ∈ B} = {γ ∈  | ξ (γ ) ∈ B} is an event. In order to make the model more close to reality, liquidity
and diversification are added to our uncertain portfolio selection
Definition 1 (Liu [27]). An uncertain variable is called zigzag if it model. Liquidity is the ability to convert investment into cash, and
has a zigzag uncertainty distribution the asset liquidity may be reflected by turnover rate [40], which
⎧ refers to the frequency of stock trading in the market within a

⎨ 0, if x < a
certain period of time. The investor usually prefer the assets with
( x − a )/2 ( b − a ), if a ≤ x < b
(x ) = good liquidity. It is acknowledged that turnover rates are hard to

⎩ ( x + c − 2b)/2(c − b), if b ≤ x < c
forecast accurately. Then we use uncertain variables to portray the
1, if x ≥ c,
turnover rates, which can be written as follows:
denoted by Z (a, b, c ), where a, b, c are real numbers with a < b < 
n
c. E[ γi xi ] ≥ κ , (3.3)
i=1
Theorem 1 (Liu [27]). Let ξ and η be independent uncertain vari-
ables with finite expected values. Then for any real numbers a and b, where γi denotes the turnover rate of the ith asset and κ repre-
we have sents the lower bound of turnover rate. Because investment di-
versification can attain the decentralized investment, we employ
E[aξ + bη] = aE[ξ ] + bE[η]. (2.1) it to reduce investment risk. Here, entropy is used to describe the
diversification of investment portfolio, which can be expressed as
Theorem 2. (Liu [27]) Let ξ and η be independent uncertain vari-
follows:
ables with regular uncertainty distributions  and , respectively.
Assume there exist two real numbers a and b such that 
n
En = − xi ln(xi + ε ) ≥ β , (3.4)
−1 (α ) = a −1
(α ) + b. i=1

for all α ∈ (0, 1 ). Show that where β is the lower bound of diversification, ε is a sufficiently
small positive number.
V [ξ + η ] = V [ξ ] + V [η]. (2.2) Obviously, investment return maximization and risk minimiza-
tion is a bi-objective optimization problem. By introducing a
Theorem 3 (Liu [27]). Let ξ be an uncertain variable with uncer-
weighting factor λ, it can be transformed into a single objective
tainty distribution  and finite expected value e . Then
optimization model. However, it can be followed by certain short-
1 coming that the different units of mean and variance can lead to
V [ξ ] = (−1 (α ) − e )2 dα . dimensional influence. In order to acquire a better accuracy, the
0

2
B. Li and R. Zhang Chaos, Solitons and Fractals 146 (2021) 110842

normalization method will be used. Based on Markowitz’s mean- 


n

variance model, we add the constraints of information entropy and xi = 1,


turnover rate to describe the diversification degree of investment i=1

and liquidity of assets. Then the uncertain mean-variance-entropy xi ≥ 0, i = 1, 2, . . . , n.


portfolio model is established as:
The model (3.7) can be established as:
 E

− E[ ni=1 ξi xi ]

V [ ni=1 ξi xi ] − Vmin

min (1 − λ ) max +λ (3.5) min ( x 1 σ1 + x 2 σ2 + . . . + x n σn ) 2 (4.2)
Emax Vmin

n

n
s.t. E[ γi xi ] ≥ κ ,
s.t. E[ γi xi ] ≥ κ ,
i=1
i=1

n

n
− xi ln(xi + ε ) ≥ β ,
− xi ln(xi + ε ) ≥ β ,
i=1
i=1

n

n
xi = 1,
xi = 1,
i=1
i=1
xi ≥ 0, i = 1, 2, . . . , n.
xi ≥ 0, i = 1, 2, . . . , n,
λ ∈ ( 0, 1 ). Proof. It follows from Theorems 1 and 2 that

In this model, Emax represents the maximum portfolio return with- E[x1 ξ1 + x2 ξ2 + . . . + xn ξn ] = x1 E[ξ1 ] + x2 E[ξ2 ] + . . . + xn E[ξn ].
out considering investment risk and Vmin is employed as the min-
and
imum portfolio risk without considering investment return. For
obtaining the optimal investment proportion of model (3.5), we V [ x 1 ξ1 + x 2 ξ2 + . . . + x n ξn ]
should first calculate Emax and Vmin under the same constraints in
= x1 V [ ξ1 ] + x 2 V [ ξ2 ] + x 3 V [ ξ3 ] + . . . + x n V [ξn ].
two auxiliary models, which are established as follows:
Since the return rate of the ith asset is a normal uncertain vari-
max E[ξ1 x1 + ξ2 x2 + . . . + ξn xn ] (3.6)
able ξi ∼ N (μi , σi ), i = 1, 2, . . . , n, the objective functions (3.6) and

n
(3.7) can be converted into the following forms:
s.t. E[ γi xi ] ≥ κ ,
i=1 max x1 μ1 + x2 μ2 + . . . + xn μn

n
and
− xi ln(xi + ε ) ≥ β ,
i=1 min(x1 σ1 + x2 σ2 + . . . + xn σn )2 .

n
The theorem is proved. 
xi = 1,
i=1 Theorem 7. Suppose the return rate of the ith asset is a zigzag
xi ≥ 0, i = 1, 2, . . . , n. uncertain variable ξi ∼ Z (ai , bi , ci ), i = 1, 2, . . . , n., then the model
(3.6) can be transformed into the following form:
and

n
ai + 2 bi + ci
min V [ξ1 x1 + ξ2 x2 + . . . + ξn xn ] (3.7) max (4.3)
4

n i=1
s.t. E[ γi xi ] ≥ κ , 
n
i=1 s.t. E[ γi xi ] ≥ κ ,

n i=1
− xi ln(xi + ε ) ≥ β , 
n
i=1 − xi ln(xi + ε ) ≥ β ,

n i=1
xi = 1, 
n
i=1 xi = 1,
xi ≥ 0, i = 1, 2, . . . , n. i=1

xi ≥ 0, i = 1, 2, . . . , n.
4. Model transformation
The model (3.7) can be transformed into the following form:
n 

In order to solve the model (3.5), we will transform two auxil- 3(bi − ai )2 + 7(ci − bi )2 + 6(bi − ai )(ci − bi )
iary uncertain optimization models (3.6) and (3.7) into two equiv- min
96
alent models. i=1

[2bi − ci − ai )+ ]3 [ ( ai + ci − 2 bi )+ ]3 2

Theorem 6. Suppose the return rate of ith asset is a normal uncertain + − xi (4.4)
variable ξi ∼ N (μi , σi ), i = 1, 2, . . . , n. the model (3.6) can be trans-
384(bi − ai ) 384(ci − bi )
formed into the following form: 
n
s.t. E[ γi xi ] ≥ κ ,
max x 1 μ1 + x 2 μ2 + . . . + x n μn (4.1) i=1

n 
n
s.t. E[ γi xi ] ≥ κ , − xi ln(xi + ε ) ≥ β ,
i=1 i=1

n 
n
− xi ln(xi + ε ) ≥ β , xi = 1,
i=1 i=1

3
B. Li and R. Zhang Chaos, Solitons and Fractals 146 (2021) 110842

Table 1
Distributions of normal uncertain return rates.

Assets i Uncertain return rate ξi Assets i Uncertain return rate ξi

1 N (0.027, 0.14 ) 6 N (0.030, 0.08 )


2 N (0.033, 0.19 ) 7 N (0.032, 0.018 )
3 N (0.032, 0.16 ) 8 N (0.031, 0.055 )
4 N (0.031, 0.15 ) 9 N (0.025, 0.10 )
5 N (0.028, 0.15 ) 10 N (0.026, 0.06 )

Table 2
Turnover rates of 10 assets.

Assets i Turnover rate γi

1 (0.011, 0.019, 0.009, 0.018 )


2 (0.009, 0.024, 0.008, 0.025 )
3 (0.007, 0.016, 0.0064, 0.018 )
4 (0.013, 0.032, 0.011, 0.030 )
5 (0.008, 0.015, 0.0076, 0.017 )
6 (0.011, 0.026, 0.009, 0.028 ) Fig. 5.1. Fmincon’s iterative convergence graph of Emax .
7 (0.012, 0.031, 0.011, 0.030 )
8 (0.006, 0.031, 0.005, 0.026 )
9 (0.011, 0.05, 0.0095, 0.047 )
10 (0.008, 0.025, 0.006, 0.026 )

xi ≥ 0, i = 1, 2, . . . , n.
Proof. Since ξi = (ai , bi , ci ) is a zigzag uncertain variable, we have

n 
n 
n 
n 
ξi x i = xi ai , xi bi , xi ci ,
i=1 i=1 i=1 i=1

which is also a zigzag uncertain variable. Hence,



n 
n
ai + 2 bi + ci
E[ x i ξi ] = ,
4
i=1 i=1

n n 
 3(bi − ai )2 + 7(ci − bi )2 + 6(bi − ai )(ci − bi )
V[ x i ξi ] =
96
i=1 i=1

[2bi − ci − ai )+ ]3 [ ( ai + ci − 2 bi )+ ]3 2
 Fig. 5.2. Fmincon’s iterative convergence graph of Vmin .
+ − xi .
384(bi − ai ) 384(ci − bi )
The theorem is proved. 
we obtain Emax =0.0357 and Vmin =0.00125. Simultaneously, the it-
5. Numerical simulation erative convergence graphs of Emax and Vmin are shown in Figs. 5.1
and 5.2. It can be found that the fmincon algorithm converges very
In this section, a numerical simulation is carried out to illus- fast. Moreover, these two iterations do not stop until the number
trate the effectiveness and practicality of the proposed uncertain of iterations reaches 39 and 42, which indicates that the fmincon
model, and to show the effect of the diversification and liquid- algorithm is an effective approach to solve the two proposed mod-
ity on portfolio selection. The following results are programmed in els. Then we fix the value of β as 1.5 and adopt five different val-
Matlab 2016a and fmincon algorithm is used. ues of λ to describe different investor preferences, like risk evader,
Suppose that the turnover rate of the ith asset is an uncertain risk enthusiast and risk rational person. The optimal portfolios un-
variable γi = (hi , li , mi , ni ) with the following uncertainty distribu- der five different values of λ are shown in Table 3. And, the cor-
tion responding objective function value (OFV), expected return E (x∗ )
⎧ and variance V (x∗ ) are also calculated. We know that the larger
⎪ 0, if x < hi
⎪ the value of λ, the lower the investor’s tolerance of investment risk
⎨ ( x − hi ) /2 ( li − hi ) ,
if hi ≤ x < li
and the lower the expectation of portfolio return, which is consis-
i (x ) = 1/2, if li ≤ x < mi (5.1)


tent with the results in Table 3.
⎩1/2 + (x − mi )/2(ni − mi ), if mi ≤ x < ni
By changing the lower bound β of diversification under λ = 0.5,
1, if x ≥ ni ,
we can analysis the influence of portfolio diversification degree on
where hi , li , mi , ni are real numbers with hi < li < mi < ni . Ob- the model (3.5). It is known that β is a parameter in the constraint
viously, the turnover rate of portfolio x = (x1 , x2 , . . . , xn ) can of diversification and the values of Emax and Vmin will change un-
n n
be described as i=1 γi xi = i=1 xi (hi , li , mi , ni ). In

addition, der different β . Then it brings about a problem that the value of
the asset liquidity can be measured by L(xi ) = E[ ni=1 γi xi ] = objective function (3.5) is hard to be estimated. For the sake of
 
E[ ni=1 xi (hi , li , mi , ni )] = ni=1 (hi , li , mi , ni )/4. Assume that asset solving this problem, we should unify the values of Emax and Vmin .
returns are normal uncertain variables. The data is obtained According to Table 4, we acquire the values are Emax = 0.030322
through the paper Huang [43]. The distributions of asset returns and Vmin = 0.001258, respectively. In Table 4, it can be seen that
and turnover rates are shown in Tables 1 and 2, respectively. the objective function value (OFV) possesses an upward trend as
The lower bound of turnover rate and diversification are set as the value of β increases, which can be interpreted as a larger value
κ =0.025 and β =1.5, respectively. According to Theorems 6 and 7, of β causes a stricter constraint condition. So it will be more dif-

4
B. Li and R. Zhang Chaos, Solitons and Fractals 146 (2021) 110842

Table 3
The optimal investment proportions for different λ under β = 1.5.

λ Optimal investment strategy x∗ E ( x∗ ) V ( x∗ ) OF V

0.1 x1 = 0.0159, x2 = 0.1089, x3 = 0.0808, x4 = 0.5705, x5 = 0.0588 0.0356 0.0139 0.0915


x6 = 0.0220, x7 = 0.0430, x8 = 0.0801, x9 = 0.0081, x10 = 0.0114
0.3 x1 = 0.0614, x2 = 0.0400, x3 = 0.0548, x4 = 0.0421, x5 = 0.0606 0.0286 0.001268 0.1315
x6 = 0.0552, x7 = 0.2068, x8 = 0.0433, x9 = 0.1122, x10 = 0.3230
0.5 x1 = 0.0629, x2 = 0.0370, x3 = 0.0515, x4 = 0.0360, x5 = 0.0578 0.0284 0.001258 0.1049
x6 = 0.0555, x7 = 0.2008, x8 = 0.0407, x9 = 0.1197, x10 = 0.3375
0.7 x1 = 0.0635, x2 = 0.0357, x3 = 0.0500, x4 = 0.0333, x5 = 0.0566 0.02834 0.001256 0.0371
x6 = 0.0557, x7 = 0.1981, x8 = 0.0395, x9 = 0.1231, x10 = 0.3440
0.9 x1 = 0.0639, x2 = 0.0350, x3 = 0.0493, x4 = 0.0318, x5 = 0.0560 0.02830 0.001255 0.0112
x6 = 0.0557, x7 = 0.1966, x8 = 0.0389, x9 = 0.1249, x10 = 0.3475

Table 4
The optimal investment proportions for different β under λ = 0.5.

β Optimal investment strategy x∗ Emax Vmin OF V

1.5 x1 = 0.0629, x2 = 0.0370, x3 = 0.0515, x4 = 0.0360, x5 = 0.0578 0.0284 0.001258 0.1049


x6 = 0.0555, x7 = 0.2008, x8 = 0.0407, x9 = 0.1197, x10 = 0.3375
2.0 x1 = 0.0663, x2 = 0.0401, x3 = 0.0548, x4 = 0.0389, x5 = 0.0612 0.0285 0.001265 0.1063
x6 = 0.0590, x7 = 0.1963, x8 = 0.0439, x9 = 0.1212, x10 = 0.3178
2.5 x1 = 0.0964, x2 = 0.0981, x3 = 0.0995, x4 = 0.1010, x5 = 0.1009 0.030309 0.002183 0.4485
x6 = 0.1009, x7 = 0.1009, x8 = 0.1009, x9 = 0.1001, x10 = 0.1009
3.0 x1 = 0.1006, x2 = 0.1006, x3 = 0.1006, x4 = 0.1005, x5 = 0.0982 0.030301 0.002185 0.4496
x6 = 0.0983, x7 = 0.1002, x8 = 0.0994, x9 = 0.1006, x10 = 0.1006
3.5 x1 = 0.1012, x2 = 0.1012, x3 = 0.1012, x4 = 0.1011, x5 = 0.0968 0.030322 0.0022 0.4562
x6 = 0.0979, x7 = 0.1011, x8 = 0.1011, x9 = 0.1012, x10 = 0.0968

ficult to obtained the optimal value of the objective function. That Acknowledgment
is, the entropy does affect the optimal investment proportions.
This work is supported by the Natural Science Foundation of
Jiangsu Province (No. BK20190787).
6. Conclusion

In this paper, we discussed an uncertain portfolio selection References


problem with considering asset liquidity and diversification degree
of portfolio, where asset returns are regarded as uncertain vari- [1] Markowitz H. Portfolio selection. J Finance 1952;7(1):77–91.
[2] Konno H, Yamazaki H. Mean absolute portfolio optimization model and its ap-
ables. First, an uncertain mean-variance-entropy model was formu- plication to Tokyo stock market. Manag Sci 1990;37(5):519–31.
lated. Different from the previous portfolio optimization model, a [3] Hogan WW, Warren JM. Toward the development of an equilibrium capital–
single objective optimization model was constructed by introduc- market model based on semivariance. J Financ Quant Anal 1974;9(1):1–11.
[4] Speranza MG. Linear programming models for portfolio optimization. Finance
ing a risk aversion factor and the dimensional influence caused by 1993;14:107–23.
different units was eliminate by normalization method. In addition, [5] Perold AF. Large-scale portfolio optimization. Manag Sci 1984;30(10):
several deterministic equivalents of the model were provided. Fi- 1143–1160.
[6] Simaan Y. Estimation risk in portfolio selection: the mean variance model ver-
nally, we carried out a numerical simulation to show the effective- sus the mean absolute deviation model. Manag Sci 1997;43(10):1437–46.
ness and practicality of our model. The results certified that diver- [7] Witbooi PJ, Schalkwyk GJV, Muller GE. An optimal investment strategy in bank
sification degree of portfolio and asset liquidity had a effect on the management. Math Method Appl Sci 2011;34(13):1606–17.
[8] Dai Z, Wang F. Sparse and robust mean-variance portfolio optimization prob-
optimal investment strategies.
lems. Phys A 2019;523:1371–8.
For the future research, we will consider the multi-period un- [9] Aksarayli M, Pala O. A polynomial goal programming model for portfo-
certain portfolio optimization models with liquidity and diversifi- lio optimization based on entropy and higher moments. Expert Syst Appl
cation and introduce some realistic constraints, such as transaction 2018;94:185–92.
[10] Bera AK, Park SY. Optimal portfolio diversification using the maximum entropy
costs, trading size and background risk. Also, the portfolio opti- principle. Econom Rev 2018;27(4-6):484–512.
mization problems with both uncertainty and randomness will be [11] Kapur JN, Kesavan HK. Entropy optimization principles with applications. New
studied. York: Academic Press; 1992.
[12] Jana P, Roy TK, Mazumder SK. Multi-objective possibilistic model for portfolio
selection with transaction cost. J Comput Appl Math 2009;228(1):188–96.
[13] Kapur JN. Maximum entropy models in science and engineering. New York:
Declaration of Competing Interest John Wiley and Sons; 1989.
[14] Zadeh LA. Fuzzy sets. Inf Control 1965;8:338–53.
[15] Deng X, Li R. A portfolio selection model with borrowing constraint based on
The authors declare that they have no known competing finan- possibility theory. Appl Soft Comput 2012;12(2):754–8.
cial interests or personal relationships that could have appeared to [16] Liu Y, Gao C. Mean-WCVar fuzzy portfolio optimization model of risk property
influence the work reported in this paper. combination. Chin J Manag Sci 2006;14(6):16–21.
[17] Zhang P, Zhang W. Multiperiod mean absolute deviation fuzzy portfolio se-
lection model with risk control and cardinality constraints. Fuzzy Sets Syst
2014;255(2):74–91.
CRediT authorship contribution statement [18] Li X, Qin Z, Kar K. Mean-variance-skewness model for portfolio selection with
fuzzy returns. Eur J Oper Res 2010;202(1):239–47.
Bo Li: Conceptualization, Methodology, Writing - original draft, [19] Vercher E, Bermúdez JD, Segura JV. Fuzzy portfolio optimization under down-
side risk measures. Fuzzy Sets Syst 2007;158(7):769–82.
Funding acquisition. Ranran Zhang: Software, Investigation, Data [20] Ammar E, Khalifa HA. Fuzzy portfolio optimization a quadratic programming
curation. approach. Chaos Solitons Fractals 2003;18(5):1045–54.

5
B. Li and R. Zhang Chaos, Solitons and Fractals 146 (2021) 110842

[21] Zhou R, Liu X. Properties of risk measures of generalized entropy in portfolio [33] Chen W, Li D, Liu Y. A novel hybrid ICA-FA algorithm for multi-period uncer-
selection. Entropy 2017;19(12):657. tain portfolio optimization model based on multiple criteria. IEEE Trans Fuzzy
[22] Mehlawat MK. Credibilistic mean-entropy models for multi-period portfolio Syst 2019;27(5):1023–36.
selection with multi-choice aspiration levels. Inf Sci 2016;345:9–26. [34] Li B, Sun Y, Aw G, Teo KL. Uncertain portfolio optimization problem under a
[23] Fang S, Rajasekera JR, Tsao HSJ. Entropy optimization and mathematical pro- minimax risk measure. Appl Math Model 2019;76:274–81.
gramming. Dordrecht: Kluwer Academic Publisheres; 1997. [35] Lu S, Zhang N, Qiu Y, Gao Y. A multi-period regret minimization model for
[24] Fang Y, Lai KK, Wang SY. Portfolio rebalancing model with transaction costs uncertain portfolio selection with bankruptcy constraint. J Intell Fuzzy Syst
based on fuzzy decision theory. Eur J Oper Res 2006;175(2):879–93. 2019;37(6):8417–39.
[25] Parra MA, Terol AB, Uria MVR. A fuzzy goal programming approach to portfolio [36] Huang X, Di H. Uncertain portfolio selection with mental accounts. Int J Syst
selection. Eur J Oper Res 2001;133(2):287–97. Sci 2020;51(12):2079–90.
[26] Huang X, Ying H. Risk index based models for portfolio adjusting problem with [37] Mehralizade R, Amini M, Gildeh BS, Ahmadzade H. Uncertain random portfolio
returns subject to experts’ evaluations. Econ Model 2013;30(30):61–6. selection based on risk curve. Soft Comput 2020;24:13331–45.
[27] Liu B. Uncertainty theory. 2nd ed. Berlin: Springer; 2007. [38] Bhattacharyya R, Chatterjee A, Kar S. Uncertainty theory based multiple objec-
[28] Huang X. Mean-variance models for portfolio selection subject to experts’ es- tive mean-entropy-skewness stock portfolio selection model with transaction
timations. Expert Syst Appl 2012;39(5):5887–93. costs. J Uncertain Anal Appl 2013;1(1):1–17.
[29] Zhang Q, Huang X, Zhang C. A mean-risk index model for uncertain capital [39] Gao J, Liu H. A risk-free protection index model for portfolio selection with
budgeting. J Oper Res Soc 2015;66(5):761–70. entropy constraint under an uncertainty framework. Entropy 2017;19:80.
[30] Qin Z, Kar S, Zheng H. Uncertain portfolio adjusting model using semiabsolute [40] Zhai J, Bai M. Mean-variance model for portfolio optimization with background
deviation. Soft Comput 2016;20(2):717–25. risk based on uncertainty theory. Int J Gen Syst 2017;47(4):1–19.
[31] Zhang P. Multiperiod mean absolute deviation uncertain portfolio selection. [41] Liu B. Some research problems in uncertainty theory. J Uncertain Syst
Soft Comput 2016;15(1):1–18. 2009;3:3–10.
[32] Li B, Zhu Y, Sun Y, Aw G, Teo KL. Multi-period portfolio selection problem [43] Huang X. Mean-risk model for uncertain portfolio selection. Fuzzy Optim Decis
under uncertain environment with bankruptcy constraint. Appl Math Model Mak 2011;10(1):71–89.
2018;56:539–50.

You might also like