Alexandria Engineering Journal (2020) 59, 739–752

H O S T E D BY
Alexandria University

Alexandria Engineering Journal

www.elsevier.com/locate/aej
www.sciencedirect.com

ORIGINAL ARTICLE

Approximate solution for solving nonlinear

fractional order smoking model
A.M.S. Mahdy a,b,*, N.H. Sweilam c, M. Higazy a,d

a
Department of Mathematics and Statistics, Faculty of Science, Taif University, Saudi Arabia
b
Department of Mathematics, Faculty of Science, Zagazig University, Zagazig, Egypt
c
Department of Mathematics, Faculty of Science, Cairo University, Giza, Egypt
d
Department of Physics and Engineering Mathematics, Faculty of Electronic Engineering, Menoufia University, Menouf, Egypt

Received 22 October 2019; revised 21 January 2020; accepted 28 January 2020

Available online 21 February 2020

KEYWORDS Abstract In this paper, Generalized Mittag-Leffler function method (GMLFM) and Sumudu
Fractional order smoking transform method (STM) are applied to study and solve the fractional order smoking model, where
model; the derivatives are defined in the Caputo fractional sense. The disease free equilibrium (DFE) and
Caputo derivative; stability of equilibrium point are studied. Lyapunov exponents and Poincare map of the model are
Generalized Mittag-Leffler drawn to ensure the stability of the model. The obtained results show that the proposed methods are
function method; very effective and convenient. In addition, the model is presented by its signal flow graph and sim-
Sumudu transform method; ulated using Matlab/Simulink.
The disease free equilibrium; Ó 2020 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria
Signal flow graph; University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
System Simulink; licenses/by-nc-nd/4.0/).
Phase space;
Stability of equilibrium
point;

MATHEMATICS SUB-
JECT CLASSIFICATION:;
41A15;
65D07;
65M12;
65K10;
68U20

1. Preliminary

As noted in [3], it was the first time to apply the notion of the
mathematical models for studying the diffusion of the disease,
which was considered the beginning of the discovery of current
* Corresponding author.
epidemiology and then followed by many scientists the discov-
E-mail address: [email protected] (A.M.S. Mahdy).
ery of diseases and epidemics [6–8,11–13,27–31]. One of the
Peer review under responsibility of Faculty of Engineering, Alexandria
most important models is the smoking model, which is the
University.
https://doi.org/10.1016/j.aej.2020.01.049
1110-0168 Ó 2020 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
740 A.M.S. Mahdy et al.

most common model for researchers [2,4,5]. The World Health

Organization’s report about the tobacco epidemic [1] states
that there are several drawbacks of the smoking such as: killing
or disables many youngs, that deprive society their basic
builders, puts high costs upon families more than it can pay,
increases the cost of health care and blocks the development
[3]. Smoking harms all parts of the body [1–4]. In [29], utilizing
the fixed point postulate, the existence of the solution has been
studied. In addition, the uniqueness of the solution is also ver-
ified via the Caputo-Fabrizio derivative definitions.
GMLFM [14–17] and STM will be applied in this paper to
study the behavior of the following the fractional order smok-
ing model [4,5]:
8 a
>
> D PðtÞ ¼ bNðtÞ  b1 LðtÞPðtÞ  ðd1 þ lÞPðtÞ þ sQðtÞ;
>
>
> a
< LðtÞ ¼ b1 LðtÞPðtÞ  b2 LðtÞSðtÞ  ðd2 þ lÞLðtÞ;
> D
Da SðtÞ ¼ b2 LðtÞSðtÞ  ðc þ d3 þ lÞSðtÞ;
>
>
>
> Da QðtÞ ¼ cSðtÞ  ðs þ d4 þ lÞQðtÞ;
>
> Fig. 1 Studied system signal flow graph.
: a
D NðtÞ ¼ ðb  lÞNðtÞ  ðd1 PðtÞ þ d2 LðtÞ þ d3 SðtÞ þ d4 QðtÞÞ;
ð1Þ 2 3
P 1 1 0 0 0 0 1
with the following initial conditions: LP 6
61 0 1 0 0 0 077
6 7
PðoÞ ¼ a1 ; LðoÞ ¼ a2 ; SðoÞ ¼ a3 ; QðoÞ ¼ a4 ; NðoÞ ¼ a5 : L 6
60 1 1 1 0 0 177
! 6 7
ð2Þ Að G Þ ¼ LS 6 0 0 1 0 1 0 0 7;
6 7
where Da is the Caputo fractional derivative operator, S 6
60 0 0 1 1 1 177
6 7
ai ; i ¼ 1; :::; 5 are constants with 0 < a 6 1: The variables Q 40 0 0 0 0 1 15
and the parameters in (1) can be defined as follows: N 1 0 0 0 0 0 1
!
The symbol The meaning Að G Þ has the following eigen values 1.4056, 2.6347, 0.6670,
1.6466 + 0.3440i, 1.6466–0.3440i, 0.5724 + 0.3187i, 0.5724–
PðtÞ potential smokers
0.3187i. Moreover, it has the following eigen vectors:
LðtÞ light smokers 2 3 2 3 2 3
SðtÞ smokers 0:2787 0:3448 0:3404
QðtÞ quit smokers 6 7 6 7 6 7
6 7 6 7 6 7
NðtÞ total smokers at time t 6 0:5546 7 6 0:3527 7 6 0:3632 7
6 7 6 7 6 7
b birth rate 6 7 6 7 6 7
l
6 7 6 7 6 7
natural birth rate 6 0:5008 7 6 0:5845 7 6 0:0981 7
c 6 7 6 7 6 7
the recovery rate from smoking 6 7 6 7 6 7
b1 ; b2 6 7 6 7 6 7
transmission coefficients 6 7 6 7 6 7
d1 ; d2 ; d3 ; d4 death rates of individuals v1 ¼ 6 0:5344 7; v2 ¼ 6 0:3918 7; v3 ¼ 6 0:7309 7;
6 7 6 7 6 7
PðtÞ; LðtÞ; SðtÞ; QðtÞ; NðtÞ related to smoking 6 7 6 7 6 7
6 7 6 7 6 7
disease 6 0:2503 7 6 0:4477 7 6 0:3894 7
6 7 6 7 6 7
s the rate at which a quite smoker in the population 6 7 6 7 6 7
6 7 6 7 6 7
becomes a potential smoker again 6 0:0482 7 6 0:1291 7 6 0:1225 7
6 7 6 7 6 7
4 5 4 5 4 5
0:1159 0:2110 0:2042
The initial conditions of model (1) must be separate of each 2 3
other and obey the rule 0:1043 þ 0:1414i
MðtÞ ¼ PðtÞ þ LðtÞ þ SðtÞ þ QðtÞ þ NðtÞ such that M is the 6 7
6 7
number of total population. 6 0:0810  0:1818i 7
6 7
Fig. 1 shows the interactions between the system’s states 6 7
6 7
! ! ! ! 6 0:0334  0:4686i 7
using its signal flow graph G ðvð G Þ; Eð G ÞÞ (where vð G Þ is 6 7
! 6 7
the set of vertices of the graph and Eð G Þ is the set of its edges). 6 7
6 7
v4 ¼ 6 0:2989  0:3470i 7;
Each vertex v correspond to a state variable and labelled
 ! by its 6 7
  6 7
symbol in model (1). So, the number of vertices is vð G Þ ¼ 7. 6 7
6 0:5781 þ 0:0000i 7
! ! 6 7
6 7
In G , ða; bÞ 2 Eð G Þ if the state corresponding to vertex a 6 7
6 0:1100 þ 0:3086i 7
affects directly the state corresponding to vertex b in model 6 7
! 4 5
(1). The following matrix is the adjacency matrix of G .
0:0351 þ 0:2373i
Approximate solutions for solving nonlinear fractional order smoking model 741
2 3 2 3
0:1043  0:1414i 0:0537  0:1801i 
Z1
6 7 6 7 f ðuÞ ¼ S½fðtÞ ¼ fðutÞet dt; where u 2 ðs1 ; s2 Þ: ð5Þ
6 0:0810 þ 0:1818i 7 6 0:3170  0:1165i 7
6 7 6 7
6 0:0334 þ 0:4686i 7 6 0:1649 þ 0:2145i 7 0
6 7 6 7
6 7 6 7 Some special properties of the sumudu transform are given
v5 ¼ 6
6 0:2989 þ 0:3470i 7 ; v
7 6 6 ¼ 6 0:1733  0:1333i 7;
7
6 7 6 7 as follows [10,19–25]:
6 0:5781 þ 0:0000i 7 6 0:2216  0:3460i 7
6 7 6 7  n1 
6 0:1100  0:3086i 7 6 0:6608 þ 0:0000i 7
4 5 4 5 t
1: S½1 ¼ 1; 2: S½t ¼ u 3: S ¼ un1 ; n > 0
0:0351  0:2373i 0:2826 þ 0:2106i CðnÞ
2 3
0:0537 þ 0:1801i
6 7 Definition 3. As defined in [10,19–25] and [30], the Sumudu
6 0:3170 þ 0:1165i 7 transform of Caputo fractional derivative is defined by the fol-
6 7
6 0:1649  0:2145i 7 lowing formula:
6 7
6 7
v7 ¼ 6
6 0:1733 þ 0:1333i 7
7;   Xm1
6 7 S Dat fðtÞ ¼ ua S½fðtÞ  uaþk fðkÞ ð0Þ; m  1 < a 6 m:
6 0:2216 þ 0:3460i 7 k¼0
6 7
6 0:6608 þ 0:0000i 7 ð6Þ
4 5
0:2826  0:2106i
Theorem ([10,19-25]).
which help in studying the topological structure of model (1). " #
h i X
n1
For more details about signal flow graph, see for example [26] ðnÞ n
S f ðtÞ ¼ u FðuÞ  u f ð0Þ ;
k k
for n P 1: ð7Þ
and [35] the references therein. k¼0
The paper is structured in 7 sections. We give some neces-
sary definitions of Caputo fractional derivative and Sumudu 1
For n ¼ 1; . . . . . . S½F0 ðtÞ ¼ ½FðuÞ  Fð0Þ: ð8Þ
transform in Section 2. The description of GMLFM is pre- u
sented in Section 3. In Section 4, the fixed point and the stabil- For more details on Sumudu transform definition and its
ity for the fractional order smoking model are studied. In properties see [10,19–25,30].
Section 5, we present an application using GMLFM for solv-
ing fractional order smoking model and we give the numerical
implementation to show the efficiency of using GMLFM and 3. The method of Mittag-Leffler
STM. In Section 6, we construct a simulation of the studied
system by the Simulink. Finally, we give the conclusions in It is well known that the power series is used to define the
Section 7. Mittag-Leffler (1902–1905) functions Ea and Ea;b [14,18]:
X
1
yr X
1
yr
2. Basic definitions and Sumudu transform Ea ðyÞ ¼ ; Ea;b ðyÞ ¼ ; a; b > 0:
r¼0
Cðra þ 1Þ r¼0
Cðra þ bÞ
In the current section, the basic definitions and properties of ð9Þ
the Sumudu transform is presented and it will be used later on.
The Mittag-Leffler functions are very useful elements in
the theory of fractional calculus and its applications since
Definition 1. The following form [9–13,18,34] gives the defini- it is already proved that it is eligible to use it as a solu-
tion of the Caputo fractional derivative Dc of order c: tion of fractional order differential and integral equations
Z [18].
1 z
fðmÞ ðtÞ In the current research, we show how to use the GMLFM
Dc fðzÞ ¼ dt; c > 0; z > 0: ð3Þ
Cðm  cÞ 0 ðz  tÞcmþ1 to solve the system of non-linear fractional order smoking
model. Using GMLFM means that zi ðyÞ, i = 1,2,3,. . .. can
Such that the symbol Cð:Þ represent the Gamma function
be decomposed to an infinite series of components [15,16] as
andm  1 < c 6 m;m 2 N:
follows:
The reader may consult [9–13,18] for more details on X
1
yra
fractional derivatives definitions and its properties. zi ðyÞ ¼ Ea ðqi ya Þ ¼ qri ; i ¼ 1; 2; 3; :::: ð10Þ
r¼0
Cðra þ 1Þ
Definition 2. As defined in [10,19–25] and [30], the Sumudu Moreover,
transform which is defined over the set of functions is given
X
1
yðr1Þa
by the following formula: Da zi ðyÞ ¼ Ea ðqi ya Þ ¼ qri ; i ¼ 1; 2; 3; ::::
Cððr  1Þa þ 1Þ
  r¼1
 jtj
ð11Þ
A ¼ fðtÞ9M; s1 ; s2 > 0; jfðtÞj < Mesj ; if t 2 ð1Þj
 That is established on the Caputo fractional derivatives.
½0; 1Þ; where j ¼ 1; 2 ; ð4Þ The reader can consult [14–17] for more details on the conver-
gence of the Mittag-Leffler function.
by the following formula:
742 A.M.S. Mahdy et al.

4. Fixed points and stability of the fractional order smoking

model

4.1. Fixed points of give up smoking system

The following equations are solved in order to determine the

equilibrium points:
Da PðtÞ ¼ bNðtÞ  b1 LðtÞPðtÞ  ðd1 þ lÞPðtÞ þ sQðtÞ ¼ 0;
ðf1Þ

Da LðtÞ ¼ b1 LðtÞPðtÞ  b2 LðtÞSðtÞ  ðd2 þ lÞLðtÞ ¼ 0; ðf2Þ

Da SðtÞ ¼ b2 LðtÞSðtÞ  ðc þ d3 þ lÞSðtÞ ¼ 0; ðf3Þ Fig. 3 Poincare map of the smoking model.

Da QðtÞ ¼ cSðtÞ  ðs þ d4 þ lÞQðtÞ ¼ 0; ðf4Þ To calculate the eigen-values, we write:

 
 ðd1 þ lÞ  k 0 0 s b 
 
Da NðtÞ ¼ ðb  lÞNðtÞ  ðd1 PðtÞ þ d2 LðtÞ þ d3 SðtÞ þ d4 QðtÞÞ ¼ 0:  0
  ðd2 þ lÞ  k 0 0 0 
 
ðf5Þ  0 0  ðc þ d3 þ lÞ  k 0 0 ¼0
 
 0 0 c  ðs þ d4 þ lÞ  k 0 

By solving (f1)–(f5) using Maple 12 with the parameters  
d1  d2  d3  d4 blk
d1 ¼ 0:33; d3 ¼ 0:55; d2 ¼ 0:44; d4 ¼ 0:66; b1 ¼ 0:01;
b2 ¼ 0:001; c ¼ 0:99; s ¼ 0:2; l ¼ 0:05; b ¼ 0:1; the model Then we have:
has three fixed points, two infeasible solutions
k1 ¼ l; k2 ¼ ðd2 þ lÞ; k3 ¼ ðc þ d3 þ lÞ;
ðP; L; S; Q; NÞ ¼ ð4:2; 1590; 531:9; 578:7; 475:3Þ,
ð49; 35:2; 0; 0; 13:8Þ and one feasible solution k4 ¼ ðs þ d4 þ lÞ; k5 ¼ ðb  l  d1 Þ
ð0; 0; 0; 0; 0Þ, for more details on equilibrium points and sta- All eigenvalues are negative that implies the stability of this
bility of nonlinear fractional models see [7,32] and the refer- fixed point and we ensure its stability by plotting its Lyapunov
ences cited there. exponents (LE1, LE2, LE3, LE4, LE5). The alogrithm
employed for determining Lyapunov exponents was proposed
4.2. Studying the stability of the trivial fixed point in [33], see Fig. 2. From Fig. 2, we see that all Lyapunov expo-
E0 ¼ ð0; 0; 0; 0; 0Þ nents are negative after small transient time that imply the sys-
tem is stable and approach its fixed point. Also, we ensure its
We calculate the Jacobian matrix as: stability by plotting the Poincare map of the model as shown in
2 3 Fig. 3. From Fig. 3, it is clear that all states of the system go to
b1 L  ðd1 þ lÞ  b1 P 0 s b
6 7 its fixed point.
6 b1 L b1 P  b2 S  ðd2 þ lÞ  b2 L 0 07
6 7
J¼6
6 0 b2 S b2 L  ðc þ d3 þ lÞ 0 0 7
7
6 7 5. Application of GMLFM
4 0 0 c  ðs þ d4 þ lÞ 0 5
d1  d2  d3  d4 bl
In this section, we will show how to solve the fractional order
Stability of E0 : smoking model (1) by using GMLFM [14]:
2 3
ðd1 þ lÞ 0 0 s b Let us consider the following
6 7
6 0 ðd2 þ lÞ 0 0 0 7
6 7 P
1 na P
1 na P
1 na
JðE0 Þ ¼ 6
6 0 0 ðc þ d3 þ lÞ 0 0 7
7 P¼ y
an Cðnaþ1Þ ; L¼ y
cn Cðnaþ1Þ ; S¼ y
en Cðnaþ1Þ ;
6 7
4 0 0 c ðs þ d4 þ lÞ 0 5 n¼0 n¼0 n¼0

d d2 d3 d4 bl P1

yna P1
yna
Q¼ f n
Cðraþ1Þ
; N¼ gn
Cðnaþ1Þ
:
n¼0 n¼0

ð12Þ
By using Caputo fraction derivative we have,
P
1 ðn1Þa P
1 na
Da PðyÞ ¼ y
an Cððn1Þaþ1Þ ¼ y
anþ1 Cðnaþ1Þ ;
n¼1 n¼0

a
P1
yðn1Þa P
1 na
D LðyÞ ¼ c n
Cððn1Þaþ1Þ
¼ y
cnþ1 Cðnaþ1Þ ;
n¼1 n¼0

a
P
1
yðn1Þa P
1 na
D SðyÞ ¼ e n
Cððn1Þaþ1Þ
¼ y
enþ1 Cðnaþ1Þ ; ð13Þ
n¼1 n¼0
P
1 ðn1Þa P1 na
Da QðyÞ ¼ fn Cððn1Þaþ1Þ
y
¼ fnþ1 Cðnaþ1Þ
y
;
n¼1 n¼0
P1 ðn1Þa P
1 na
Da NðyÞ ¼ y
gn Cððn1Þaþ1Þ ¼ y
gnþ1 Cðnaþ1Þ :
n¼1 n¼0
Fig. 2 Dynamics of Lyapunov exponents for smoking model.
Approximate solutions for solving nonlinear fractional order smoking model 743

Substituting Eqs. (12) and (13) in system (1) we claim In system (16), yna is impossible equal to zero then the coef-
P
1 P
1 P
1 ficients are equal to zero and we get the recurrence relations
na na
y
anþ1 Cðnaþ1Þ ¼b y
gn Cðnaþ1Þ  b1 rn yna which helps to calculate the constants values
n¼0 n¼0 n¼0
an ; cn ; en ; fn ; gn ; n ¼ 1; 2; 3; :::; 1:
P
1 na P
1 na
ðd1 þ lÞ y
an Cðnaþ1Þ þs fn Cðnaþ1Þ
y
; anþ1 ¼ bgn  b1 rn Cðna þ 1Þ  ðd1 þ lÞan þ sfn ;
n¼0 n¼0
P
1 na P
1 P
1 cnþ1 ¼ b1 rn Cðna þ 1Þ  b2 rn1 Cðna þ 1Þ  ðd2 þ lÞcn ;
y
cnþ1 Cðnaþ1Þ ¼ b1 rn yna  b2 rn1 yna
n¼0 n¼0 n¼0 enþ1 ¼ b2 rn1 Cðna þ 1Þ  ðc þ d3 þ lÞen ; ð17Þ
P
1
yna fnþ1 ¼ cen  ðc þ d4 þ lÞfn ;
ðd2 þ lÞ cn Cðnaþ1Þ ;
n¼0 gnþ1 ¼ ðb  lÞgn  d1 an  d2 cn  d3 en  d4 fn :
P
1 na P
1 P
1 na
y
enþ1 Cðnaþ1Þ ¼ b2 rn1 yna  ðc þ d3 þ lÞ y
en Cðnaþ1Þ ; At n = 1
n¼0 n¼0 n¼0
P1 na P
1 na P1 na a ¼ bg1  b1 r1 Cða þ 1Þ  ðd1 þ lÞa1 þ sf1 ;
2
fnþ1 Cðnaþ1Þ
y
¼c y
en Cðnaþ1Þ  ðc þ d4 þ lÞ fn Cðnaþ1Þ
y
;
n¼0 n¼0 n¼0 c2 ¼ b1 r1 Cða þ 1Þ  b2 r11 Cða þ 1Þ  ðd2 þ lÞc1 ;
P1
yna P
1
yna e2 ¼ b2 r11 Cða þ 1Þ  ðc þ d3 þ lÞe1 ; ð18Þ
gnþ1
Cðnaþ1Þ
¼ ðb  l Þ g n
Cðnaþ1Þ
n¼0 n¼0 f ¼ ce  ðc þ d4 þ lÞf ;
2 1 1

P
1 P1
d1 an yna
Cðnaþ1Þ
 d2 cn yna
Cðnaþ1Þ
g2 ¼ ðb  lÞg1  d1 a1  d2 c1  d3 e1  d4 f1 ;
n¼0 n¼0
1 1 1 1 1 1 0
P1 na P1 na where r1 ¼ Cðaþ1Þ
a c
þ Cðaþ1Þ
r
; r11 ¼ Cðaþ1Þ
c e
þ Cðaþ1Þ
c e
.
d3 y
en Cðnaþ1Þ  d4 fn Cðnaþ1Þ
y
;
n¼0 n¼0 Similarly, we can find a3 ; c3 ; e3 ; f3 ; g3 . . . . . .
Substituting (15), (18) into (12), we have the solution in the
P
n P
n
infinite series as follows:
where rn ¼ an cnk
Cðkaþ1ÞCððnkÞaþ1Þ
; rn1 ¼ cn enk
Cðkaþ1ÞCððnkÞaþ1Þ
.
k¼0 k¼0 1 a 2 2a 3 3a

Then after compilation the summations we derive the P ¼ a0 þ Cðaþ1Þ

a y
þ Cð2aþ1Þ
a y
þ Cð3aþ1Þ
a y
þ ;
1 a
following L ¼ c0 þ Cðaþ1Þ
c y
þ Cð2aþ1Þ
c y 2 2a
þ Cð3aþ1Þ
c y 3 3a
þ ;
1 n
P o 1 a
þ b1 rn þ ðCðnaþ1Þ
d1 þlÞan 2 2a 3 3a
bgn sfn
anþ1
Cðnaþ1Þ
 Cðnaþ1Þ  Cðnaþ1Þ yna ¼ 0 S ¼ e0 þ Cðaþ1Þ
e y
þ Cð2aþ1Þ
e y
þ Cð3aþ1Þ
e y
þ ; ð19Þ
n¼0
P1 n o 0 f1 ya f2 y2a f3 y3a
Q ¼ f þ Cðaþ1Þ þ Cð2aþ1Þ þ Cð3aþ1Þ þ    ;
cnþ1
Cðnaþ1Þ
 b1 rn þ b2 rn1 þ ðCðnaþ1Þ
d2 þlÞcn
yna ¼ 0; 1 a 2 2a 3 3a
n¼0 N ¼ g0 þ Cðaþ1Þ
g y
þ Cð2aþ1Þ
g y
þ Cð3aþ1Þ
g y
þ :
P1 n o
 b2 rn1 þ ðcþd 3 þlÞe
n
enþ1
Cðnaþ1Þ Cðnaþ1Þ
yna ¼ 0;
n¼0
P1 n n
o 5.1. Numerical implementation using GMLFM
fnþ1
Cðnaþ1Þ
cen
 Cðnaþ1Þ þ ðcþd 4 þlÞf
Cðnaþ1Þ
yna ¼ 0;
n¼0
P1 n o In this section, we implement GMLFM to get numerically
gnþ1 ðblÞgn d1 an cn d 2 en d 3 fn d4
Cðnaþ1Þ
 Cðnaþ1Þ þ Cðnaþ1Þ þ Cðnaþ1Þ þ Cðnaþ1Þ þ Cðnaþ1Þ yna ¼ 0: results of the fractional order smoking model of Eqs. (1) with:
n¼0

ð14Þ a1 ¼ 45; a2 ¼ 65; a3 ¼ 60; a4 ¼ 80; a5 ¼ 250;

As system (14) non-homogenous then we find the first term d1 ¼ 0:33; d2 ¼ 0:44; d3 ¼ 0:55;
by creating the first limit of the summation and get:
a1 ¼ bg0  b1 r0  ðd1 þ lÞa0 þ sf0 ; r0 ¼ a0 c0 :
c1 ¼ b1 r0  b2 r01  ðd2 þ lÞc0 ; r01 ¼ c0 e0 :
e1 ¼ b2 r01  ðc þ d3 þ lÞe0 : ð15Þ
f1 ¼ ce0  ðc þ d4 þ lÞf0 :
g1 ¼ ðb  lÞg0  d1 a0  d2 c0  d3 e0  d4 f0 :
Then system (14) become
1 n
P o
anþ1
Cðnaþ1Þ
bgn
 Cðnaþ1Þ þ b1 rn þ ðCðnaþ1Þ
d1 þlÞan sfn
 Cðnaþ1Þ yna ¼ 0
n¼1
P1 n o
cnþ1
Cðnaþ1Þ
 b1 rn þ b2 rn1 þ ðCðnaþ1Þ
d2 þlÞcn
yna ¼ 0;
n¼1
P1 n o
 b2 rn1 þ ðcþd 3 þlÞe
n
enþ1
Cðnaþ1Þ Cðnaþ1Þ
yna ¼ 0;
n¼1
P1 n n
o
fnþ1
Cðnaþ1Þ
cen
 Cðnaþ1Þ þ ðcþd 4 þlÞf
Cðnaþ1Þ
yna ¼ 0;
n¼1
P1 n o
gnþ1 ðblÞgn d1 an cn d 2 en d 3 fn d4
Cðnaþ1Þ
 Cðnaþ1Þ þ Cðnaþ1Þ þ Cðnaþ1Þ þ Cðnaþ1Þ þ Cðnaþ1Þ yna ¼ 0:
n¼1
Fig. 4 The behavior of approximate solution of P(t) with
ð16Þ a = 0.85.
744 A.M.S. Mahdy et al.

Fig. 7 The behavior of approximate solution of Q(t) with

Fig. 5 The behavior of approximate solution of L(t) with a = 0.85.
a = 0.85.

Fig. 8 The behavior of approximate solution of N(t) with

a = 0.85.

Figs. 4–15 show the calculation further terms of PðtÞ;LðtÞ,

Fig. 6 The behavior of approximate solution of S(t) with SðtÞ; QðtÞ and NðtÞ of the approximate solution of fractional
a = 0.85. order smoking model by using GMLFM with
a ¼ 0:85 and a ¼ 0:999.
d4 ¼ 0:66; b1 ¼ 0:01; b2 ¼ 0:001; c ¼ 0:99; s ¼ 0:2;
5.2. Application using STM
l ¼ 0:05; b ¼ 0:1;
In this section, we implement STM to the fractional order
PðtÞ ¼ 45 þ 2:53804 t0:85 - 2:85236 t1:7
smoking model of Eqs. (1):
Taking the Sumudu transform to both sides of Eqs. (1), we
LðtÞ ¼ 65 - 10:1522 t0:85 þ 4:52131 t1:74
get
8
SðtÞ ¼ 60 - 98:3491 t0:85 þ 92:9471 t1:76 >
> S½Da PðtÞ ¼ S½bNðtÞ  b1 LðtÞPðtÞ  ðd1 þ lÞPðtÞ þ sQðtÞ;
>
>
> a
< S½D LðtÞ ¼ ½b1 LðtÞPðtÞ  b2 LðtÞSðtÞ  ðd2 þ lÞLðtÞ;
>
QðtÞ ¼ 80 - 14:1707 t0:85 - 51:7102 t1:78 S½Da SðtÞ ¼ ½b2 LðtÞSðtÞ  ðc þ d3 þ lÞSðtÞ;
>
>
>
> S½Da QðtÞ ¼ ½cSðtÞ  ðs þ d4 þ lÞQðtÞ;
>
>
NðtÞ ¼ 250 - 187:604 t0:85 þ 84:0663 t1:72 :
S½Da NðtÞ ¼ ½ðb  lÞNðtÞ  ðd1 PðtÞ þ d2 LðtÞ þ d3 SðtÞ þ d4 QðtÞÞ:
ð20Þ
Approximate solutions for solving nonlinear fractional order smoking model 745

Fig. 9 The behavior of approximate solution of P(t), L(t), S(t),

Fig. 12 The behavior of approximate solution of S(t) with
Q(t) and N(t) with a = 0.85.
a = 0.999.

Fig. 10 The behavior of approximate solution of P(t) with

a = 0.999.
Fig. 13 The behavior of approximate solution of Q(t) with
a = 0.999.

Using the property of the Sumudu transform and the initial

condition in Eq. (20), we have:
8
>
> S½PðtÞ ¼ Pð0Þ þ ua :S½bNðtÞ  b1 LðtÞPðtÞ  ðd1 þ lÞPðtÞ þ sQðtÞ
>
>
> a
< ½LðtÞ ¼ Lð0Þ þ u :½b1 LðtÞPðtÞ  b2 LðtÞSðtÞ  ðd2 þ lÞLðtÞ;
> S
S½SðtÞ ¼ Sð0Þ þ ua :½b2 LðtÞSðtÞ  ðc þ d3 þ lÞSðtÞ;
>
>
>
> S½QðtÞ ¼ Qð0Þ þ ua :½cSðtÞ  ðs þ d4 þ lÞQðtÞ;
>
>
:
S½NðtÞ ¼ Nð0Þ þ ua :½ðb  lÞNðtÞ  ðd1 PðtÞ þ d2 LðtÞ þ d3 SðtÞ þ d4 QðtÞÞ

8
>
> S½PðtÞ ¼ 20 þ ua :S½bNðtÞ  b1 LðtÞPðtÞ  ðd1 þ lÞPðtÞ þ sQðtÞ;
>
>
> a
< ½LðtÞ ¼ 40 þ u :½b1 LðtÞPðtÞ  b2 LðtÞSðtÞ  ðd2 þ lÞLðtÞ;
> S
S½SðtÞ ¼ 60 þ ua :½b2 LðtÞSðtÞ  ðc þ d3 þ lÞSðtÞ;
>
>
>
> S½QðtÞ ¼ 80 þ ua :½cSðtÞ  ðs þ d4 þ lÞQðtÞ;
>
>
:
S½NðtÞ ¼ 200 þ ua :½ðb  lÞNðtÞ  ðd1 PðtÞ þ d2 LðtÞ þ d3 SðtÞ þ d4 QðtÞÞ:
ð21Þ

Operating with the Sumudu inverse on both sides of Eq.

Fig. 11 The behavior of approximate solution of L(t) with (21) we get
a = 0.999.
746 A.M.S. Mahdy et al.
X1 h h X1 X1 ii
n¼0
Sn ðtÞ ¼ 60 þ S1 ua : b2 n¼0 Bn ðtÞ  ðc þ d3 þ lÞ n¼0 Sn ðtÞ ;

X1 h h X1 X1 ii
n¼0
Qn ðtÞ ¼ 80 þ S1 ua : c n¼0 Sn ðtÞ  ðs þ d4 þ lÞ n¼0 Qn ðtÞ ;

X1
Nn ðtÞ ¼ 200 þ S1
 
n¼0
P1 P1 
X1 d1 n¼0 Pn ðtÞ þ d2 n¼0 Ln ðtÞ
ua : ðb  lÞ n¼0 Nn ðtÞ  P P1 :
þd3 1n¼0 Sn ðtÞ þ d4 n¼0 Qn ðtÞ

where An ; Bn are Adomian polynomials that represent nonlin-

ear term. So Adomian polynomials are given as follows:
An ðtÞ ¼ LðtÞPðtÞ; Bn ðtÞ ¼ LðtÞSðtÞ
The few components of the Adomian polynomials are given
as follows:
A0 ¼ L0 ðtÞP0 ðtÞ; A1 ¼ L0 ðtÞP1 ðtÞ þ L1 ðtÞP0 ðtÞ;
A2 ¼ L0 ðtÞP2 ðtÞ þ L1 ðtÞP1 ðtÞ þ L2 ðtÞP0 ðtÞ;
Fig. 14 The behavior of approximate solution of N(t) with B0 ¼ L0 ðtÞS0 ðtÞ; B1 ¼ L0 ðtÞS1 ðtÞ þ L1 ðtÞS0 ðtÞ;
a = 0.999. B2 ¼ L0 ðtÞS2 ðtÞ þ L1 ðtÞS1 ðtÞ þ L2 ðtÞS0 ðtÞ:
Then we have:
P0 ¼ 20; L0 ¼ 40; S0 ¼ 60; Q0 ¼ 80; N0 ¼ 200;
A0 ¼ 800; B0 ¼ 2400:

Pkþ1 ðtÞ ¼ S1 ½ua :S½bNk ðtÞ  b1 Ak ðtÞ  ðd1 þ lÞPk ðtÞ þ sQk ðtÞ;
ð24Þ

Lkþ1 ðtÞ ¼ S1 ½ua :½b1 Ak ðtÞ  b2 Bk ðtÞ  ðd2 þ lÞLk ðtÞ;

Skþ1 ðtÞ ¼ S1 ½ua :½b2 Bk ðtÞ  ðc þ d3 þ lÞSk ðtÞ;

Qkþ1 ðtÞ ¼ S1 ½ua :½cSk ðtÞ  ðs þ d4 þ lÞQk ðtÞ;

Nkþ1 ðtÞ ¼ S1 ½ua :½ðb  lÞNk ðtÞ  ðd1 Pk ðtÞ þ d2 Lk ðtÞ þ d3 Sk ðtÞ þ d4 Qk ðtÞÞ

P1 ðtÞ ¼ S1 ½ua :S½bN0 ðtÞ  b1 A0 ðtÞ  ðd1 þ lÞP0 ðtÞ þ sQ0 ðtÞ;

L1 ðtÞ ¼ S1 ½ua :½b1 A0 ðtÞ  b2 B0 ðtÞ  ðd2 þ lÞL0 ðtÞ;

Fig. 15 The behavior of approximate solution of P(t), L(t), S(t),
Q(t) and N(t) with a = 0.999.
8
S1 ðtÞ ¼ S1 ½ua :½b2 B0 ðtÞ  ðc þ d3 þ lÞS0 ðtÞ;
>
> PðtÞ ¼ 20 þ S1 ½ua :S½bNðtÞ  b1 LðtÞPðtÞ  ðd1 þ lÞPðtÞ þ sQðtÞ;
>
>
> 1 a
< LðtÞ ¼ 40 þ S ½u :½b1 LðtÞPðtÞ  b2 LðtÞSðtÞ  ðd2 þ lÞLðtÞ;
> Q1 ðtÞ ¼ S1 ½ua :½cS0 ðtÞ  ðs þ d4 þ lÞQ0 ðtÞ;
1 a
SðtÞ ¼ 60 þ S ½u :½b2 LðtÞSðtÞ  ðc þ d3 þ lÞSðtÞ;
>
> N1 ðtÞ ¼ S1 ½ua :½ðb  lÞN0 ðtÞ  ðd1 P0 ðtÞ þ d2 L0 ðtÞ þ d3 S0 ðtÞ þ d4 Q0 ðtÞÞ
>
> QðtÞ ¼ 80 þ S1 ½ua :½cSðtÞ  ðs þ d4 þ lÞQðtÞ;
>
>
:
NðtÞ ¼ 200 þ S1 ½ua :½ðb  lÞNðtÞ  ðd1 PðtÞ þ d2 LðtÞ þ d3 SðtÞ þ d4 QðtÞÞ:
ð22Þ P1 ðtÞ ¼ Cð20:4
aþ1Þ
ta ; L1 ðtÞ ¼ Cð14
aþ1Þ
ta ; S1 ðtÞ ¼ Cð93
aþ1Þ
ta ; Q1 ðtÞ ¼ C13:4
ðaþ1Þ
ta ;

By assuming that N1 ðtÞ ¼ C100

ðaþ1Þ
ta ; A1 ¼ Cð536
aþ1Þ
ta ; B1 ¼ C4560
ðaþ1Þ
ta :
P P P1
P ðt Þ ¼ 1n¼0 Pn ðtÞ; LðtÞ ¼ 1 n¼0 Ln ðtÞ; SðtÞ ¼ n¼0 Sn ðtÞ; By similarity:
P P1
Q ðt Þ ¼ 1 Q ð tÞ; N ð tÞ ¼ N ð t Þ
n¼0 n n¼0 n
25:792 2a 16:78 2a
ð23Þ P2 ðtÞ ¼ t ; L2 ðtÞ ¼ t ;
Cð2a þ 1Þ Cð2a þ 1Þ
By substituting Eq. (23) in Eq. (22) we have 143:31 2a 79:876 2a
P1   P P1 P1
S2 ðtÞ ¼ t ; Q 2 ðt Þ ¼ t ;
1 a
u :S b 1 Cð2a þ 1Þ Cð2a þ 1Þ
n¼0 Pn ðtÞ ¼ 20 þ S n¼0 Nn ðtÞ  b1 n¼0 An ðtÞ  ðd1 þ lÞ n¼0 Pn ðtÞ
P1 
þs n¼0 Qn ðtÞ ;
P1   P1 P P1  54:872 2a 696:08 2a 285:6
n¼0 Ln ðtÞ ¼ 40 þ S
1 a
u : b1 n¼0 An ðtÞ  b2 1n¼0 Bn ðtÞ  ðd2 þ lÞ n¼0 Ln ðtÞ ; N2 ðtÞ ¼ t ; A2 ¼ t  t2a
Cð2a þ 1Þ Cð2a þ 1Þ Cða þ 1Þ:Cða þ 1Þ
Approximate solutions for solving nonlinear fractional order smoking model 747

Fig. 16 System Simulation by MATLAB\Simulink.

 
4:158Cð2a þ 1Þ t3a
L3 ðtÞ ¼ 21:9222  ;
Cða þ 1Þ:Cða þ 1Þ Cð3a þ 1Þ
 
1:302Cð2a þ 1Þ t3a
S3 ðtÞ ¼ 221:1237 þ ;
Cða þ 1Þ:Cða þ 1Þ Cð3a þ 1Þ

t3a
Q3 ðtÞ ¼ 214:56406 ;
Cð3a þ 1Þ

t3a
N3 ðtÞ ¼ 22:23058 ;
Cð3a þ 1Þ
Since
Fig. 17 The plot represents the dynamics of the population of PðtÞ ¼ P0 þ P1 þ P2 þ P3 þ   
potential smokers with respect to time in weeks.
LðtÞ ¼ L0 þ L1 þ L2 þ L3 þ   

6739:2 2a 1302 SðtÞ ¼ L0 þ S1 þ S2 þ S3 þ   

B2 ¼ t þ t2a
Cð2a þ 1Þ Cða þ 1Þ:Cða þ 1Þ
QðtÞ ¼ Q0 þ Q1 þ Q2 þ Q3 þ   
And:
 
2:856Cð2a þ 1Þ t3a NðtÞ ¼ N0 þ N1 þ N2 þ N3 þ   
P3 ðtÞ ¼ 6:27376 þ ;
Cða þ 1Þ:Cða þ 1Þ Cð3a þ 1Þ Then
748 A.M.S. Mahdy et al.

Fig. 18 The plot represents the dynamics of the population of occasional smokers with respect to time in weeks.

Fig. 19 The plot represents the dynamics of the population of smokers with respect to time in weeks.

Fig. 20 The plot represents the dynamics of the population of quite smokers with respect to time in weeks.
Approximate solutions for solving nonlinear fractional order smoking model 749

Fig. 21 The plot represents the dynamics of the total population of smokers with respect to time in weeks.

13:4 a 79:876 2a t3a

QðtÞ ¼ 80  t  t þ 214:56406
Cða þ 1Þ Cð2a þ 1Þ Cð3a þ 1Þ
þ 

100 54:872 2a t3a

NðtÞ ¼ 200  ta þ t  22:23058
Cða þ 1Þ Cð2a þ 1Þ Cð3a þ 1Þ
þ 

6. System simulation

In this section, a simulation of the system is constructed using

Simulink given in Fig. 16. Referring to Figs. 17–21, it is obvious
that the outputs of the studied system via the suggested methods
completely agree with that from the Simulink. Furthermore, the
interaction and dependency of the components of the system on
each other are clear from its Simulink simulation.
Figs. 17–21, show the simulation responses that completely
agree with the analytical solution.

6.1. Phase space

Fig. 22 represents the phase space.

Fig. 22 The plot represents the phase space. 6.2. Parameters space

Figs. 23 and 24 represent the Parameters space.

20:4 a 25:792 2a
PðtÞ ¼ 20 þ t  t 7. Conclusions
Cða þ 1Þ Cð2a þ 1Þ
 
2:856Cð2a þ 1Þ t3a
þ 6:27376 þ þ  In this paper, a mathematical model of the nonlinear fractional
Cða þ 1Þ:Cða þ 1Þ Cð3a þ 1Þ
order smoking problem are studied which plays a vital role in
14 16:78 2a applied sciences. Also, the generalized Mittag-Leffler function
LðtÞ ¼ 40  ta þ t method and Sumudu transform method are applied to get the
Cða þ 1Þ Cð2a þ 1Þ
  approximate solutions of this model. The signal flow graph of
4:158Cð2a þ 1Þ t3a the studied system is constructed. The disease-free equilibrium
 21:9222 þ þ 
Cða þ 1Þ:Cða þ 1Þ Cð3a þ 1Þ (DFE) and stability of equilibrium point are studied. Stability
of the system is ensured by plotting its Lyapunov exponents
93 143:31 2a and it is found that all Lyapunov exponents are negative after
SðtÞ ¼ 60  ta þ t
Cða þ 1Þ Cð2a þ 1Þ small transient time. Also, the stability is ensured by the
 
1:302Cð2a þ 1Þ t3a Poincare map of the model. We present a simulation of the sys-
 221:1237  þ  tem using Simulink of MATLAB. The phase space and Param-
Cða þ 1Þ:Cða þ 1Þ Cð3a þ 1Þ
eter space are studied.
750 A.M.S. Mahdy et al.

Fig. 23 The plot represents the Parameters space.

Approximate solutions for solving nonlinear fractional order smoking model 751

Fig. 24 The plot represents the Parameters space.

752 A.M.S. Mahdy et al.

8. Authors’ contributions [15] A.A.M. Arafa, S.Z. Rida, H.M. Ali, Generalized Mittag-Leffler
function method for solving Lorenz system, Int. J. Innov. Appl.
Stud. 3 (2013) 105–111.
This paper equally between the authors contributed. The final [16] A.A.M. Arafa, S.Z. Rida, A.A. Mohammadein, H.M. Ali,
manuscript approved and read between the authors. Solving nonlinear fractional differential equation by generalized
Mittag-Leffler function method, Commun. Theor. Phys. J. 59
Declaration of Competing Interest (2013) 661–663.
[17] S.Z. Rida, A.A.M. Arafa, New method for solving linear
The authors declare that they have no known competing fractional differential equations, Int. J. Differential Equations
financial interests or personal relationships that could have 2011 (2011) 1–8.
[18] I. Podlubny, Fractional Differential Equations, Academic Press,
appeared to influence the work reported in this paper.
London, 1999.
[19] N. Bildik, S. Deniz, The use of Sumudu decomposition method
Acknowledgments for solving predator-prey systems, Int. J. Math. Sci. Lett. 3
(2016) 285–289.
The authors are grateful to the editor and the referees for care- [20] S.Z. Rida, A.S. Abedl-Rady, A.A.M. Arafa, H.R. Abedl-
fully reading the paper and for their comments and suggestions Rahim, A domian decomposition Sumudu transform method
which have improved the paper. for solving fractional nonlinear equations, Math. Sci. Lett. 5 (1)
(2016) 39–48.
[21] S. Momani, Z. Odibat, Numerical approach to differential
References equations of fractional order, J. Comput. Appl. Math. (2006).
[22] S. Deniz, N. Bildik, Comparison of Adomian decomposition
[1] A.H. Mokdad, J.S. Marks, D.F. Stroup, J.L. Gerberding, method and taylor matrix method in solving different kinds of
Actual causes of death in the United States, JAMA 291 (2004) partial differential equations, Int. J. Model. Optim. 4 (4) (2014)
1238–1245. 292–298.
[2] O.K. Ham, Stages and processes of smoking cessation among [23] S. Rathore, D. Kumar, J. Singh, S. Gupta, Homotopy analysis
adolescents, West J. Nurs. Res. 29 (2007) 301–315. sumudu transform method for nonlinear equations, Int. J. Ind.
[3] O.D. Makinde, Adomian decomposition approach to a SIR Math. 4 (4) (2012).
epidemic model with constant vaccination strategy, Appl. Math. [24] J. Singh, D. Kumar, Homotopy perturbation sumudu transform
Comput. 184 (2017) 842–848. method for nonlinear equations, Adv. Theor. Appl. Mech. 4 (4)
[4] J. Singh, D. Kumar, M. Al-Qurashi, D. Baleanu, A new (2011) 165–175.
fractional model for giving up smoking dynamics, Adva. Diffe. [25] A.M.S. Mahdy, A.S. Mohamed, A.A.H. Mtawa,
Equa. 2017 (2017) 1–16. Implementation of the Homotopy perturbation Sumudu
[5] F. Haq, K. Shah, G. Rahman, M. Shahzad, Numerical solution transform method for solving Klein-Gordon equation, Appl.
of fractional order smoking model via laplace Adomian Math. 6 (3) (2015) 617–628.
decomposition method, Alexandria Eng. J. 57 (2) (2018) 1061– [26] H. Schmid, A. Huber, Analysis of switched-capacitor circuits
1069. using driving-point signal-flowgraphs, AnalogIntegr Circ Sig
[6] A.M.A. El-Sayed, S.M. Salman, On a discretization process of Process 96 (2018) 495–507.
fractional order Riccati’s differential equation, J. Fract. Calc. [27] A.M.S. Mahdy, M. Higazy, Numerical different methods for
Appl. 4 (2013) 251–259. solving the nonlinear biochemical reaction model, Int. J. Appl.
[7] A.A. Elsadany, A.E. Matouk, Dynamical behaviors of Comput. Math. 5 (6) (2019) 1–17.
fractional-order Lotka-Voltera predator-prey model and its [28] K.A. Gepreel, A.M.S. Mahdy, M.S. Mohamed, A. Al-Amiri,
discretization, Appl. Math. Comput. 49 (2015) 269–283. Reduced differential transform method for solving nonlinear
[8] M. El-Shahed, J.J. Nieto, A.M. Ahmed, I.M.E. Abdelstar, biomathematics models, CMC: Comput. Mater. Continua 61 (3)
Fractional-order model for biocontrol of the lesser date moth in (2019) 979–994.
palm trees and its discretization, Adv. Difference Equat. 2017 [29] J. Singh, D. Kumar, M. Al Qurashi, D. Baleanu, A new
(2017) 1–16. fractional model for giving up smoking dynamics, Adv.
[9] S. Momani, Z. Odibat, Numerical approach to differential Difference Equat., 2017, (2017), pp. 1-16. DOI 10.1186/
equations of fractional order, J. Comput. Appl. Math. 207 s13662-017-1139-9.
(2007) 96–110. [30] A. Prakash, M. Kumar, D. Baleanu, A new iterative technique
[10] H. Bulut, H.M. Baskonus, F.B.M. Belgacem, The analytical for a fractional model of nonlinear Zakharov-Kuznetsov
solutions of some fractional ordinary differential equations equations via Sumudu transform, Appl. Math. Comput. 334
by sumudu transform method, Abstr. Appl. Anal. 2013 (2013) (2018) 30–40.
1–6. [31] J. Singh, D. Kumar, D. Baleanu, S. Rathore, An efficient
[11] I. Hashim, M. Chowdhurly, S. Mawa, On multistage homotopy numerical algorithm for the fractional Drinfeld-Sokolov-Wilson
perturbation method applied to nonlinear biochemical reaction equation, Appl. Math. Comput. 335 (2018) 12–24.
model, Chaos, Solitons Fractals 36 (2008) 823–827. [32] H.A.A. El-Saka, The fractional order SIS epidemic model with
[12] Y.A. Amer, A.M.S. Mahdy, E.S.M. Youssef, Solving systems of variable population size, J. Egyptian Math. Soc. 22 (2014) 50–54.
fractional differential equations using Sumudu transform [33] A. Wolf, J.B. Swift, H.L. Swinney, J.A. Vastano, Determining
method, Asian Res. J. Math. 7 (2) (2017) 1–15. Lyapunov exponents from a time series, Phys. D Nonlinear
[13] H. Jafari, V. Daftardar-Gejji, Solving a system of nonlinear Phenom. 16 (1985) 285–317.
fractional differential equations using Adomian decomposition, [34] A.M.S. Mahdy, Numerical studies for solving fractional integro-
J. Comput. Appl. Math. 196 (2) (2006) 644–651. differential equations, J. Ocean Eng. and Sci. 3 (2) (2018) 127–132.
[14] S.Z. Rida, Y.Gh. Gouda, H.M. Ali, M.M. Farag, Approximate [35] R. Scapellato, R. El-Shanawany, M. Higazy, Orthogonal double
solution for SIR epidemic model by generalized Mittag-Leffler covers of Cayley graphs, Disc. Appl. Math. 157 (14) (2009)
function method, Int. J. Adv. Res. Sci. Eng. Technol. 5 (4) 3111–3118.
(2018) 5519–5525.