Downloaded from http://iranpaper.

ir
http://www.itrans24.com/landing1.html

Soft Computing
https://doi.org/10.1007/s00500-019-03862-2

METHODOLOGIES AND APPLICATION

Fuzzy simulation of European option pricing using mixed fractional

Brownian motion
Sara Ghasemalipour1 · Behrouz Fathi-Vajargah2

© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract
Financial pricing models have great impact on the world of high finance as they enable financial experts to predict the dynamics
of underlying asset. Over the last few decades, there has been a lot of competitions among financial researches to establish
the most efficient pricing model for different options. This study aims to propose an option valuation model based on mixed
fractional Brownian motion and to show how it can efficiently be used as a financial predictive model. In fact, this option
evaluation model employs the fuzzy simulation method to estimate a European call option under the condition that the interest
rates (domestic and foreign rates) and the volatility are random fuzzy variables. Furthermore, the performance of the proposed
model is validated by solving some experimental problems.

Keywords Mixed fractional Brownian motion · European call option · Fuzzy simulation · Random fuzzy variables

1 Introduction returns of the financial assets in the Black–Scholes approach

are assumed to be linearly distributed, while the asset prices
A reliable evaluation of option pricing is of the utmost impor- in real-world condition often show a distribution curve with
tance to both the investment companies and traders involved characteristics such as heavy tails, self-similarity, volatil-
in the world of high finance. Thus, a wide range of differ- ity clustering and long-range dependence (see Lo 1991;
ent option pricing models have been introduced by many Liu et al. 2016; Mudzimbabwe et al. 2012; Mantegna and
researches in order to capture the stochastic behavior of Stanley 1996; Willinger et al. 1999). To get around this
underlying asset price process. issue, a variant of stochastic process called fractional Brow-
Generally, the Black–Scholes model is considered as one nian motion (hereafter called fBm) was adopted by some
of the most widely used evaluation techniques for different researches to model the random behavior of underling asset
options. In this mathematical model, it is assumed that the price. Rogers (1997) in his study suggested that the fBm
input parameters, such as risk-free interest rates and volatil- not only can describe long-range correlation between mar-
ity, are constants and the fluctuation in stock prices follows ket returns associated with different days, but also enables
a geometric Brownian motion. Although the Black–Scholes arbitrage opportunities. Shortly after that, Cheridito (2003)
model is able to evaluate different types of options without reported the results of an analytical study in which an arbi-
arbitrage opportunity, its results often are not satisfactory as trage strategy was designed for a financial market comprising
compared to the reality. This is mainly because the logarithm money account and a stock whose price tends to behave as an
fBm. Also, an extensive research was carried out by Zhang
Communicated by V. Loia. and Xiao (2017) to investigate arbitrage opportunities using
various types of fractional Gaussian processes such as fBm.
B Sara Ghasemalipour
[email protected] To address some possible flaws in the fBm model, El-Nouty
Behrouz Fathi-Vajargah
(2003) and Mishura (2008) suggested a mixed form of Brow-
[email protected] nian motion and fractional Brownian motion model (called
mfBm) with long memory property as a viable financial
1 Faculty of Mathematical Sciences, University of Guilan, model for describing the random behavior of the stock assets
Rasht, Iran
from time to time. Cheridito (2001) researched into mfBm
2 Department of Statistics, Faculty of Mathematical Sciences, model and indicated that if Hurst parameter H ∈ ( 43 , 1) then
University of Guilan, Rasht, Iran

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

S. Ghasemalipour, B. Fathi-Vajargah

any form of market arbitrage is not allowed. Prakasa Rao fuzzy logic, alpha-cuts and simulation approach. The final
(2016) developed an analytical evaluation technique based section discusses validation of the proposed model which is
on an mfBm model of underlying stock price and applied it performed using some numerical experiments.
to geometric Asian power option with fixed strike right at
initial time (t = 0). 2 Definitions and preliminaries
On the other hand, since the financial market is very prone
to both volatility and uncertainty, sometimes it is impossible In this section, some basic concepts of random fuzzy vari-
to record and collect the data exactly. In other words, the ables and the mfBm are provided. Further descriptions about
most data produced by complex financial market are essen- them are available in Liu (2004, 2007), Liu and Liu (2003a,
tially linguistic instead of numerical. As a result, it is not 2002, 2003b).
feasible to exactly define the model parameters, e.g., risk-
free interest rate r and volatility σ . In fact, this model makes 2.1 Random fuzzy variables
an assumption in which the parameters are considered as
fuzzy numbers instead of the crisp numbers. Yoshida (2003) Suppose that  is a nonempty set and P represents the power
estimated the fuzzy price of European call option by model- set of  (i.e., the largest—algebra over ) in which each
ing the stochastic behavior of the underlying asset in the form element is named as an event. A self-evident definition of
of a fuzzy process. Lee et al. (2005) applied both the fuzzy credibility is expressed by assigning to each event A a num-
logic theory and Bayes rule to obtain the optimal actions ber Cr {A} that represents the credibility with which A will
for Black–Scholes model and suggested that the lack of the happen. The set function Cr is named as a credibility mea-
fuzziness notion in the modeling process of European call sure if it can meet a set of axioms including the normality,
option will result in misjudgment. Thiagarajah et al. (2007) monotonicity, self-duality and maximality. Then, the triplet
developed a financial model in which the constant parame- (, P(), Cr ) will be considered as a credibility space.
ters in the standard Black–Scholes model were replaced with Fuzzy variable A fuzzy variable can be determined by a mea-
fuzzy variables, including fuzzy interest rate, fuzzy volatility surable function from a credibility space (, P(), Cr ) to
and fuzzy stock price. The most useful feature of this model the set of real-valued numbers. Hence, it will have a member-
is to enable the financial analysts to reasonably select any ship function which is defined from the credibility measure
European and American option price for the investors who by
make decision accordingly. For more details, see (Wu 2004,
2005, 2007; Muzzioli and Reynaerts 2008; Yoshida 2002). μ(x) = (2Cr {ξ = x}) ∧ 1, x ∈ . (1)
Recently, mfBm as a reliable mathematical model is being
utilized to estimate the option prices based on probability Let ξ be a fuzzy variable defined on the credibility space
theory. However, as far as the author of this study is aware, (, P(), Cr ), and α ∈ (0, 1]. Then,
the application of mfBm is not yet investigated for evaluating
European options in terms of fuzzy sets. Hence, in this paper, ξα = inf{r |μξ (r ) ≥ α}, ξα = sup{r |μξ (r ) ≥ α}, (2)
an attempt is made to establish a financial model that can
effectively evaluate a European call option through the mfBm are called α-cuts of ξ . Then, the expected value of ξ is defined
resulting from fuzzy simulation technique. In other words, by
the proposed approach considers the risk-free interest rates  
∞ 0
and their corresponding volatilities as random fuzzy variables E[ξ ] = Cr {ξ ≥ r }dr − Cr {ξ ≤ r }dr , (3)
to find a reliable solution for this approach associated with the 0 −∞
real stock market. In the present study, the whole modeling
process is carried out in two steps: In the first step, the alpha- provided that at least one of the above two integrals is finite.
cuts of option price are estimated by applying the theory of Proposition 1 Let ξ be a fuzzy variable defined on the cred-
fuzzy sets under the assumption that the risk-free interest ibility space (, P(), Cr ). Then, we have
rates (foreign and domestic rates) and volatility are random
fuzzy variables. In the second step, the fuzzy simulation is  1
1
employed to calculate the expected value of payoff which is E[ξ ] = (ξα + ξα )dα. (4)
2 0
corresponded to the call option of European type.
A general overview on the remainder of this study is pro- In a recently published study, Liu and Liu (2003a) devel-
vided by summarizing it into the following sections: Sect. 2 oped a mathematical formula based on a random fuzzy
gives a brief introduction to the basic principles of fuzzy variable in order to describe various phenomena in which ran-
set theory and mfBm. Section 3 describes a computational domness and fuzziness co-occur in the context of optimiza-
method for call option evaluation of European type using tion. Random fuzzy variable is considered as an extended

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

Fuzzy simulation of European option pricing using mixed fractional Brownian motion

form of fuzzy variable whose realization value is a random (4). For 0 ≤ s < t, the random variable B = Bt − Bs has a
variable rather than a real number. From mathematical point normal distribution with mean 0 and variance σ 2 (t −s).
of view, a random fuzzy variable can be expressed as a func-
tion from the credibility space (, P(), Cr ) to the set of
Mandelbrot and Van Ness (1968) firstly introduced fBm
variables with random values. The expected value for a ran-
by extending the Brownian motion through an additional
dom fuzzy variable is computed from
parameter called Hurst parameter, taking values in the inter-
 val [0, 1]. However, the only problem with the fBm is that
∞
the possibility of arbitrage is not taken into account. More
E[ξ ] = Cr {ω ∈ |E[ξ(ω)] ≥ r }dr
0 specifically, fBm as a centered Gaussian process provides
 0 stationary but not independent increments, except when it
− Cr {ω ∈ |E[ξ(ω)] ≤ r }dr . (5) turns into the standard Brownian motion. To model the
−∞
dependency feature of the increments, the Hurst parameter
(0 < H < 1) is defined as satisfying the mathematical con-
2.2 An overview of Brownian motion model and its dition B0H = E[BtH ] = 0, t ≥ 0. The covariance function
extensions of fBm is calculated from

Over the last few decades, a series of empirical researches 1 2H

have investigated the importance of the Brownian motion Cov(BsH , BtH ) = (s + t 2H + (t − s)2H ), s < t.
2
in modeling the behavior of financial asset prices and sug-
gested it as a rich source of randomness and uncertainty
that can assist the financial experts in solving real-world The increment process {Bk+1 H − B H }, k = 0, 1, 2, . . . can be
k
problems. An example of such applications is the financial considered as a specific type of Gaussian process, namely
model of Black–Scholes–Merton which applies the Brown- Fractional Gaussian Noise (FGN). By carrying out straight-
ian motion to predict the dynamics of various options. On the forward calculations, it is found that
other hand, some analytical studies recently offered substan-
tial evidence against the ability of the Brownian motion to (1) BtH exhibits stationary increments.
generate randomness. The reasons for this claim are twofold. (2) BtH is self-similar, meaning that both the {B H (αt), t ≥
First, the asset returns seen in real stock markets do not fluc- 0} and {α H BtH , t ≥ 0} defined as data points in finite-
tuate according to the Gaussian law as their distributions dimensional space are identically distributed for all the
are characterized by high kurtosis and heavy tails. Second, α values greater than zero α > 0.
time series of daily returns (return values) follow a distri- (3) BtH has long-range dependency feature for 21 < H < 1,
bution with long-range dependency feature. However, the meaning that there are positive correlations between the
above-mentioned criticisms can be addressed by adopting process increments.
the following strategies. In the case of the first criticism, var-
ious classes of heavy-tailed distribution, including the stable,
tempered stable and geo-stable distributions, are proposed in To prove the possibility of arbitrage, Rogers (1997) con-
order to model the return fluctuations. It should be noted ducted an analytical study and indicated that the fBm is not
that all of the mentioned distributions are classified as Levy well able to describe time series of financial returns. For this
processes. In the case of the second criticism, the fBm is reason, he introduced mfBm as an efficient financial model
considered to correctly reflect the long-range dependency of the return values and defined it as follows: An mfBm of
attribute of the financial prices. From a mathematic perspec- three parameters α, β and H is to combine different fBms
tive, Brownian motion {Bt ; t ≥ 0} is representative of a linearly within the probability space (, F, {Ft , t ≥ 0}, P)
random, continuous-time process that captures continuous for all t ∈ R+ , as expressed below:
stochastic movement. It can be viewed either as a continuous-
time random walk or as a continuous-time Gaussian process MtH = α Bt + β BtH , (6)
whose increments are uncorrelated. The main properties of
the Brownian motion are described as follows:
where Bt is a Brownian motion, BtH is an independent frac-
tional Brownian motion of Hurst parameter H ∈ (0, 1), α and
(1). B0 = 0, that is, it initiates right at zero. β are two real constants such that (α, β) = (0, 0), {Ft , t ≥ 0}
(2). Bt is continuous at t ≥ 0, that is, it consists of contin- denotes the filtration generated by (Bt0 , BtH0 ) for t0 ≤ t.
uous sample paths in which there are no jumps. For more detailed information about the properties of the
(3). It has both stationary and independent increments. mfBm, one can refer to Xiao et al. (2012), Zili (2006).

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

S. Ghasemalipour, B. Fathi-Vajargah

3 Fuzzy option pricing and computational (8)

methods
where S̃0 is the initial stock price, r̃d , r̃ f , σ̃ are the same
This section explains how to derive a mathematical formula parameters in formula (7) and we assume that they are ran-
for evaluating call option of European type in fuzzy envi- dom fuzzy variables.
ronment. During the last decade, a wide range of scientific For financial market, some ideal conditions are postulated
studies have suggested that fBm has considerable potential as follows:
for application in finance. For instance, Cheridito (2001) and
Rogers (1997) developed financial forecasting approaches
(i) The stock price St meets the stochastic differential equa-
based on fBm process with some changes in Gaussian ker-
tion of the mixed fractional model.
nel of the integral representation. It should be noted that
(ii) Purchasing or selling (Trading) the financial assets is
modeling of asset prices can only be accomplished under
accomplished without any transaction costs or taxes,
condition that there is no arbitrage possibility. By considering
that is, the financial market has a frictionless feature.
the underlying asset process as a semimartingale, it is possi-
(iii) The Hurst exponent is set in the range of ( 43 , 1), which
ble to provide arbitrage-free opportunities. However, Rogers
means there is no arbitrage possibility in the financial
(1997) in his analytical study indicated that fBm process is
market.
not essentially a semimartingale. As an alternative to arbi-
trage elimination, Cheridito (2001) proposed the following
formula for describing the dynamics of financial asset price: One of the most challenging tasks in option valuation is find-
ing a reasonable price C(St , t), termed as contract function.
d St = (rd − r f )St dt + σ St d Bt + σ St d BtH , This implies that the option holder has to pay right at time t
when the stock is worth St . Consider a call option of Euro-
S0 > 0, 0 ≤ t ≤ T , (7)
pean type such that C̃(t, S̃t ) represents the fuzzy price at
time t and the strike price K is matured at time T . There are
where r f and rd represent domestic and foreign interest rates,
two ways to compute an option price: (1) estimate the risk-
respectively, σ is the volatility of St , Bt is a Brownian motion,
neutral expected discounted payoff. (2) numerically solve
BtH is an independent fractional Brownian motion with Hurst
a partial differential equation. Here, we address the first
parameter with H ∈ ( 43 , 1) fulfilling the free-arbitrage oppor-
of these methods. Then, by defining C̃(t, S̃t ) as a random
tunity.
fuzzy variable, the price of European call option is calcu-
Since underlying asset prices observed in international
lated from
stock markets are highly fluctuating in nature, both ran-
domness and uncertainty may occur frequently. Due to
imprecise information and the fluctuating nature of stock C̃(t, S̃t ) = E Q [e−r̃d (T −t) ( S̃T − K )+ |Ft ], (9)
prices, sometimes it is difficult to determine some input
parameters accurately for the models of interest. Thus, this where K is the strike price, T is the expiration time, +
uncertainty should be incorporated into the modeling pro- is the maximum S̃T − K and zero and E Q [.|F] denotes
cess. For this reason, fuzzy set theory is employed to express the quasi-condition expectation with respect to informa-
expert opinions or imprecision in the estimation of the market tion filtration Ft under the risk-neutral measure Q. Hence,
parameters. It is also possible to translate statistical estima- for any ω ∈ , E Q [e−r̃d (ω)(T −t) ( S̃T (ω) − K )+ ] can be
tors into fuzzy numbers. In fact, when these parameters are considered as a fuzzy variable whose membership func-
set to constant, the final simulation results will not be in good tion is μ E[ξ ] where ξ = e−r̃d (ω)(T −t) ( S̃T (ω) − K )+ . The
agreement with real-world condition, and therefore, it seems expected value ξ is calculated based on this member-
a good idea to determine them as random fuzzy variables. ship function, although sometimes it is difficult to obtain
According to what was just mentioned, the domestic and it. Details of the computational procedure for estimating
foreign interest rates, volatility and asset prices are assumed the expected value are explained in the following subsec-
to be random fuzzy variables. tions.
Suppose that (, F, P) represents a complete probability
space in which all information is filtered as {Ft }t∈[0,T ] . It is
also possible to consider mixed fractional class of the stan- 3.1 Alpha-cuts scheme
dard Black–Scholes model as a simple pricing model with a
stock price whose dynamics are described by Generally, the α-cuts are considered as the mapping functions
which are commonly used in fuzzy set theory. Based on the
 
1 2 α-cuts, it is possible to make specific fuzzy patterns for the
S̃t = S̃0 ex p (r̃d − r̃ f )t + σ̃ (Bt + Bt ) − σ̃ (t + t )
H 2H
of European call option evaluation model. Mathematically,
2

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

Fuzzy simulation of European option pricing using mixed fractional Brownian motion

the α-cuts of the solution (8) are expressed as the mathematical definition of the fBm suggested by Kroese
 and Botev (2015), a single sample path of the mfBm can eas-
   ily be obtained. To simulate the mfBm, it is first required to
St,α = S0,α ex p (rd,α − r f ,α )t + σα (Bt + BtH )
develop the mathematical models of Brownian motion and

1 fBm, respectively. By considering one dimension of Brown-
− (σ 2 )α (t + t 2H ) , (10) ian motion, it can be stated that this model is the cumulative
2
 sum of a displacement vector with normally distributed ran-
  
St,α = S0,α ex p (rd,α − r f ,α )t + σα (Bt + BtH ) dom elements. Hence, the Brownian motion model is defined
 by successively adding some mathematical terms of normally
1
− (σ 2 )α (t + t 2H ) . (11) distributed random numbers, that is
2
 , S  , r  , r  , r  , r  , σ  , σ  are the α-cuts
where S0,α B0 ∼ N (0, σ 2 )
0,α d,α d,α f ,α f ,α α α
of S̃0 , r̃d , r̃ f , σ̃ , respectively. Obviously, S̃t , function related B1 ∼ B0 + N (0, σ 2 )
to them, is a random fuzzy variable and for any ω ∈ , B2 ∼ B1 + N (0, σ 2 )
E[ S̃t (ω)] is a fuzzy variable. Therefore, the α-cuts of random ...
fuzzy price are as follows:
where N represents the normal distribution functions which
E Q [e−r̃d (ω)(T −t) ( S̃T (ω) − K )+ ]α = inf{x|μ E[ξ(ω)] (x) ≥ α}, are assumed to be independent from each other. Among var-
(12) ious approaches developed based on the mfBm, the most
efficient one is circulant method which is introduced by
and Coeurjolly (2000). In fact, the main idea behind this method
is to express a Gaussian distribution function ζ with mean μ
E Q [e−r̃d (ω)(T −t) ( S̃T (ω)−K )+ ]α = sup{x|μ E[ξ(ω)] (x) ≥ α}, and covariance matrix C as ξ = μ + Sζ in which ζ denotes
a standard Gaussian vector, and the matrix S is defined as
(13)
SS T = C. So, according to the circulant method, a square
root of covariance matrix is required to define a Gaussian vec-
Based on Eq. (4), the expected value of (9) can be estimated
tor. Therefore, in the present study, the fBm is simulated by
as
applying the above-mentioned method. After that, the Brow-
nian motion and fBm paths are combined together to generate
C̃(t, S̃t ) = E Q [e−r̃d (ω)(T −t) ( S̃T (ω) − K )+ ]
 random paths for the mfBm model. Figure 1 illustrates the
1 1  generated paths for the Brownian motion, fBm and mfBm
= (ξ + ξα )dα. (14)
2 0 α models when the Hurst parameter (H ) is equal to 0.3, 0.6,
0.8 and 0.9. According to this Figure, it is seen that there is
where ξα , ξα are the α-cuts of ξ . The obtained solution is the an increase in fluctuations when H is close to zero, and in
desired price with an acceptable accuracy. However, Eq. (14) contrast, the fluctuations are reduced when H is close to one.
needs to the α-cuts of expression ξ for being applied in prac-
tice. For this purpose, fuzzy simulation method is described
3.2.2 Fuzzy simulation method
as follows.
As mentioned before in Sect. 3, the expected value of a
3.2 Simulation approach
fuzzy variable is calculated from its membership function.
Due to difficulty in obtaining the membership function of
In the present study, the modeling process is divided into
E Q [e−r̃d (ω)(T −t) ( S̃T (ω) − K )+ ], Liu and Liu (2003a) sug-
two parts: First, the mfBm paths are generated. Second, the
gested a technique to estimate it. Hence, in the present study,
generated paths are employed in simulating the call option.
the expected value of stock price is computed based on the
above-mentioned technique in order to evaluate the proposed
3.2.1 Generating the random paths for mixed fractional
call option. This computation is carried out by defining the
Brownian motion
following algorithm.
Suppose that ξ = (r̃d , r̃ f , σ̃ , σ̃ 2 , S̃0 ) represents a fuzzy
As far as the author is aware, no program code is available
variable whose joint membership function is expressed as
in the current literature for pricing the call options based on
mfBm. Thus, in the present study, the mfBm is utilized to
develop a new financial model for call option pricing. From μξ (u) = min μi (u)
1≤i≤5

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

S. Ghasemalipour, B. Fathi-Vajargah

Fig. 1 The simulated paths of Bm, fBm, mfBm for H = 0.3, 0.6, 0.8, 0.9

where μ1 = μrd , μ2 = μr f , μ3 = μσ , μ4 = μσ 2 , μ5 = (5) Calculate W1 = 21 (μ1 + max j≥1 μ j − max j>1 μ j ) and
μ S0 .
 
1
(1) Uniformly generate u i = (u i1 , u i2 , u i3 , u i4 , u i5 ), i = Wk = max μ j − max μ j + max μ j − max μ j ,
2 1≤ j≤k 1≤ j<k k≤ j k< j
1, 2, . . . , N . k = 1, 2, . . . , N
(2) Calculate μi = min μ j (u i ), i = 1, 2, . . . , N .
1≤i≤5
i (T −t) (u i −u i )T +u i (B +B H )− 1 u i (T +T 2H ) N
(3) Set xi = u 5 e
i u 1 (e 1 2 3 T T 2 4
(6) Set E Q [e−r̃d (ω)(T −t) ( S̃T (ω) − K )+ ] =
+ i=1 x i Wi
− K) .
(4) Sort xi and μi , i = 1, 2, . . . , N .

Table 1 The price of European

n 100 1000 2000 5000 10,000 15,000 20,000 30,000 50000
call option with respected to
simulation paths H = 0.75 0.0816 0.0980 0.1098 0.1106 0.1117 0.1125 0.1145 0.1183 0.1194
H = 0.8 0.0739 0.1058 0.1162 0.1235 0.1249 0.1273 0.1313 0.1342 0.1365
H = 0.9 0.0835 0.1296 0.1351 0.1471 0.1575 0.1610 0.1654 0.1676 0.1684

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

Fuzzy simulation of European option pricing using mixed fractional Brownian motion

Table 2 The price of European call option with respect to Hurst param-
eter H
H 0.75 0.8 0.85 0.9 0.95 0.99

Price 0.1194 0.1365 0.1474 0.1684 0.2003 0.2257

 
rd,α = 0.05 − α, rd,α = 0.05 + α
r f ,α = 0.003 − α, r f ,α = 0.003 + α
σα = 0.25 − α, σα = 0.25 + α
 
S0,α = 1.5 − α, S0,α = 1.5 + α

then the α-cuts for the fuzzy term ξ are

ξα = e−(0.05+α)(T −t) ((1.5 − α)

H )− 1 (0.0625+α)(T +T 2H )
Fig. 2 The convergence of the fuzzy simulation for the different values e((0.05−α)−(0.003+α))T +(0.25−α)(BT +BT 2
of H and n
−K )+ ,

and
4 Validation of the proposed pricing model
ξα = e−(0.05−α)(T −t) ((1.5 + α)
In this section, an experimental example of call option trad- H )− 1 (0.0625−α)(T +T 2H )
ing is provided to check the performance of the proposed e((0.05+α)−(0.003−α))T +(0.25+α)(BT +BT 2

model. In fact, the accuracy of the suggested pricing model −K )+ .

is evaluated by fuzzy simulation method. For this purpose, a
European call option is considered to be estimated using the By assuming t = 0.6, K = 1.2, T = 1 and for H = 0.8 the
proposed model. Moreover, the simulated results are com- mfBm path value is −0.3767. Thus, from Eq. (14), the price
pared with the results produced by α-cut solution. of European call option is calculated equal to 0.1364.

Now, to test the efficiency of the proposed model, the input

Example 1 Consider a financial model with input parameters parameters in Example 1 are performed by fuzzy simulation
equal to r̃d = (0.04/0.05/0.06), r̃ f = (0.002/0.003/0.004), method.
σ̃ = (0.24/0.25/0.26), S̃0 = (1.4/1.5/1.6) , H = 0.8 and In the first case, the simulation number is increased and
[0, T ] and use Eq. (14) to obtain the option price C̃. The the Hurst parameter is set to 0.8, 0.85, 0.9 and 0.95, respec-
α-cuts of input parameters are follows: tively. Table 1 displays the simulation results for different

Fig. 3 The convergence of the fuzzy simulation for the different values of H and K

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

S. Ghasemalipour, B. Fathi-Vajargah

Table 3 The price of European call option with respect to strike price Ethical approval This article does not contain any studies with human
K participants or animals performed by any of the authors.
K 0.7 0.8 0.9 1 1.1 1.2

Price 0.6118 0.5410 0.3481 0.3190 0.2261 0.1365

References
simulation numbers and Hurst values. Figure 2 illustrates the
price convergence of European call option which is a func- Cheridito P (2001) Mixed fractional Brownian motion. Bernoulli
tion of the simulation number. According to this figure, it is 7(6):913–934
Cheridito P (2003) Arbitrage in fractional Brownian motion models.
seen that the real solution in Example 1 can be reached by
Finance Stoch 7(4):533–553
increasing the simulation number. After 50000 simulations, Coeurjolly J (2000) Simulation and identification of the fractional
the price value of 0.1365 will be obtained for H = 0.8, which Brownian motion: a bibliographical and comparative study. J Stat
is a reasonable solution. Softw 5(7):1–53
El-Nouty C (2003) The fractional mixed fractional Brownian motion.
In the second case, the Hurst parameter is changed in the
Stat Probab Lett 65(2):111–120
range of ( 43 , 1) while the number of simulations stays con- Kroese D, Botev Z (2015) Spatial process simulation, stochastic geom-
stant (n = 50000). Table 2 and Fig. 3 show the results of etry, spatial statistics and random fields. Springer, Berlin, pp
the simulation method. According to Fig. 3, the call option 369–404
Lee C, Tzeng CH, Wang SY (2005) A new application of fuzzy set
price is increasing with respect to the Hurst parameter H that
theory to the Black–Scholes option pricing model. Expert Syst
is considered as an acceptable result. It means that the more Appl 29(2):330–342
the Hurst index, the more is the option price value. Liu B (2004) Uncertainty theory. Springer, Berlin
In the third case, the strike price K is changed while the Liu B (2007) Uncertainty theory, 2nd edn. Springer, Berlin
Liu B, Liu Y (2002) Expected value of fuzzy variable and fuzzy expected
simulation number and the Hurst index stay fixed (i.e., n =
value models. IEEE Trans Fuzzy Syst 10(4):445–450
50000 and H = 0.8). Table 3 and Fig. 3 illustrate the results Liu Y, Liu B (2003) Expected value operator of random fuzzy variable
of the simulation. According to Fig. 3, the call option price is and random fuzzy expected value models. Int J Uncertain Fuzzi-
decreasing with respect to strike price K . The lower the strike ness Knowl Based Syst 11(2):195–215
Liu Y, Liu B (2003) Fuzzy random variables: a scalar expected value
price, the greater is the option price. Therefore, it is concluded
operator. Fuzzy Optim Decis Making 2(2):143–160
that increasing the Hurst parameter H and lowering the strike Liu Y, Zhang W, Wang J (2016) Multi-period cardinality constrained
price K can lead to price growth in the option pricing. portfolio selection models with interval coefficients. Ann Oper Res
244(2):545–569
Lo A (1991) Long-term memory in stock market prices. Econometrica
59(5):1279–1313
Mandelbrot B, Van Ness J (1968) Fractional Brownian motions, frac-
5 Conclusion tional noises and applications. Siam Rev 10(4):422–437
Mantegna R, Stanley H (1996) Turbulence and financial markets. Nature
Designing an efficient financial valuation approach can give 38:587–588
a better understanding of the dynamics of underlying assets. Mudzimbabwe W, Patidar K, Witbooi P (2012) A reliable numerical
method to price arithmetic Asian options. Appl Math Comput
In this paper, an attempt is made to develop a new mfBm- 218(22):10934–10942
based model that estimates a fuzzy European call option with Mishura Y (2008) Stochastic calculus for fractional Brownian motions
Hurst exponent H ∈ ( 43 , 1). This option evaluation model and related processes. Springer, Berlin
is constructed by considering the long memory property of Muzzioli S, Reynaerts H (2008) American option pricing with imprecise
risk-neutral probabilities. Int J Approx Reason 49(1):140–147
underlying asset price and providing the arbitrage-free oppor- Prakasa Rao B (2016) Pricing geometric Asian power options under
tunities. In fact, the whole modeling procedure is divided into mixed fractional Brownian motion environment. Physica A Phys-
two parts: First, the fuzzy European call option is modeled ica A Stat Mech Appl 446(15):92–99
based on the mfBm. Second, the fuzzy simulation method is Rogers L (1997) Arbitrage with fractional Brownian motion. Math
Finance 7(1):95–105
applied for obtaining the final option price. Finally, to vali- Thiagarajah K, Appadoo S, Thavaneswaran A (2007) Option valu-
date the model effectiveness, two experimental problems are ation model with adaptive fuzzy numbers. Comput Math Appl
considered to be solved. The obtained simulation results indi- 53(5):831–841
cate that the proposed financial model is able to estimate the Willinger W, Taqqu M, Teverovsky V (1999) Stock market prices and
long-range dependence. Finance Stoch 3(1):1–13
European call option with high degree of precision. Wu H (2004) Pricing European options based on the fuzzy pattern of
Black–Scholes formula. Comput Oper Res 31(7):1069–1081
Wu H (2005) European option pricing under fuzzy environment. Int J
Compliance with ethical standards Intell Syst 20(1):89–102
Wu H (2007) Using fuzzy sets theory and Black–Scholes formula to gen-
Conflict of interest The author declares that they have no conflict of erate pricing boundaries of European options. Appl Math Comput
interest. 185(1):136–146

123
Downloaded from http://iranpaper.ir
http://www.itrans24.com/landing1.html

Fuzzy simulation of European option pricing using mixed fractional Brownian motion

Xiao W, Zhang W, Zhang X (2012) Pricing model for equity warrants Zhang X, Xiao W (2017) Arbitrage with fractional Gaussian processes.
in a mixed fractional Brownian environment and its algorithm. Physica A Stat Mech Appl 471:620–628
Physica A Stat Mech Appl 391(24):6418–6431 Zili M (2006) On the mixed fractional Brownian motion. J Appl Math
Yoshida Y (2002) Optimal stopping models in a stochastic and fuzzy Stoch Anal 2006:1–9
environment. Inf Sci 142(1):89–101
Yoshida Y (2003) The valuation of European options in uncertain envi-
ronment. Eur J Oper Res 145(1):221–229 Publisher’s Note Springer Nature remains neutral with regard to juris-
dictional claims in published maps and institutional affiliations.

123