Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: In this paper, the problem of the spread of a non-fatal disease in a population known as a nonlinear epidemic
Homotopy perturbation method model is solved using an Optimized Homotopy Perturbation Method (OHPM). In order to discretize the gov
Particle swarm optimization erning nonlinear differential equations, the Finite Difference Method (FDM) is applied to approximate the de
Optimized homotopy perturbation method
rivatives of the problem and the penalty method is employed to satisfy the initial conditions. The unknown nodal
Epidemic model
Nonlinear differential equations
values of the discretized objective function are regarded as the design variables and would be found by the
Particle Swarm Optimization (PSO) algorithm. In order to find more accurate results, the solution proposed by
the Homotopy Perturbation Method (HPM) is considered as the global best particle of PSO at the first iteration of
the optimization process, and this idea is named the Optimized Homotopy Perturbation Method (OHPM). The
results obtained by the introduced methodology are associated with those of the HPM, PSO, and a numerical
method to assess the performance of the OHPM.
1. Introduction concluded that the large noise can make the disease die out exponen
tially, and the stationary distribution will exist under certain conditions
Infectious diseases, which are caused by pathogens, are a detriment [5]. Zhang et al. presented epidemic dynamics in an adaptive network
to the public health. It is difficult to analyze the intrinsic rules of diseases where the susceptible are able to avoid contact with the infectious by
through human trials; thus qualitative and quantitative analyses could rewiring their network connections. They depicted that such rewiring of
in theory be an important method. Hence, epidemic models, as impor the local connections changes the topology of the network, and inevi
tant tools to understand the spread and control of infectious diseases, tably has a profound effect on the transmission of the disease, which in
have attracted many researchers in the recent years. For instance, Liu turn influences the rewiring process [6]. Moreover, the occurrence of a
and Zhong studied modeling and analyzing the dynamic spreading of the sustained periodic solution via the Hopf bifurcation in an age-structured
epidemic by a network eigenvalue method and mainly focuses on epidemic model under the assumption that the transmission rate de
studying the influence of network characteristics on malware spreading pends on the age of infective individuals and the product of the trans
[1]. Fan et al. proposed two spatio-temporal epidemic network models mission rate were studied by Kuniya [7]. Besides, Zhang et al.
based on popularity and similarity optimization (PSO), in which new investigated and studied the backward bifurcation in an especial type of
connections take both popularity and similarity into account [2]. Khan the stage-structured epidemic models [8]. Buonomo and Marca explored
et al. described the formulation of a hepatitis B epidemic model with the qualitative dynamics of a vaccine-preventable disease, when the
saturated incidence rate and discussed the basic properties (existence of healthy individuals in a given community are subject to two coexisting
positive solutions and positively invariant set etc.) to show the mathe threats: the imperfection of the vaccine and the voluntary (non-)
matical as well as the biological feasibility [3]. Further, Wei et al. adherence to immunization programs [9]. Finally, a numerical method
investigated the dynamics of the stochastically perturbed Heroin based on the Magnus series expansion and its second-order convergence
epidemic model under non-degenerate noises devoted to discern long were inspected by Csomo �s to a system of quasilinear delay equations
time dynamics and proved it to have a unique global positive solution [10].
[4]. Cao et al. examined a stochastic epidemic model with quarantine Besides, a lot of analytical and numerical approaches have been
and standard incidence and consider the existence and uniqueness of the introduced to solve these kind of the nonlinear differential equations, to
global positive solution to the stochastic system. Further, they name but a few, Monotone positive approaches [11], Logistic function
https://doi.org/10.1016/j.imu.2020.100293
Received 14 November 2019; Received in revised form 12 January 2020; Accepted 12 January 2020
Available online 13 January 2020
2352-9148/© 2020 The Author. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
M.J. Mahmoodabadi Informatics in Medicine Unlocked 18 (2020) 100293
2
M.J. Mahmoodabadi Informatics in Medicine Unlocked 18 (2020) 100293
300 300
OHPM OHPM
PSO-FVM PSO-FVM
250 250
Global Best Position
150 150
100 100
50 50
0 0
0.5 1 1.5 2 0.5 1 1.5 2
Iteration x 10
4 Iteration x 10
4
(a) (b)
400 400
OHPM OHPM
350 PSO-FVM 350 PSO-FVM
300 300
Global Best Position
250 250
200 200
150 150
100 100
50 50
0 0
0.5 1 1.5 2 0.5 1 1.5 2
Iteration x 10
4
Iteration x 10
4
(c) (d)
Fig. 1. Convergence trajectories of the OHPM and PSO-FVM for (a) 5, (b) 6, (c) 9 and (d) 11 nodal points.
X
M � �
μ n j þ jbν n j þ jbξ n jg þ Pfjμn ðT1 Þ
fjb μ0 j þ jνn ðT1 Þ ν0 j þ jξn ðT1 Þ ξ0 jg ¼ 0; θ σ; ∂σ=∂n ¼ 0; ρ 2 Γ; (13)
i¼1
(11)
where η represents a general differential operator, θ dignifies a boundary
where, P≫1 is the penalty parameter and here, set at P ¼ 100. It is operator, φðρÞ characterizes a known analytical function, Γ symbolizes
noticeable that the forward FDM and backward FDM are implemented the boundary of domain Ω and ∂=∂n denotes differentiation along the
for the first and end nodal points, respectively. normal drawn outwards from Ω. Always, it is possible to divide operator
η into linear Λ and nonlinear Δ parts. Hence, Equation (12) could be
4. Homotopy perturbation method and the analytical solution of rewritten as follows:
the problem
ΛðσÞ þ ΔðσÞ φðρÞ ¼ 0: (14)
The homotopy perturbation method as a combination of the classical A homotopy of Equation (14) is constructed with εðρ; λÞ : Ω � ½0; 1�→
perturbation technique and homotopy scheme is initially introduced by R which satisfies
He in 1999 for solving nonlinear differential equations as [36]:
ψ ðε; λÞ ¼ ð1 λÞΛðεÞ Λðσ0 Þ þ λðηðεÞ φðρÞ ¼ 0; λε½0; 1�; ρ 2 Ω; (15)
ηðσ Þ φðρÞ ¼ 0; ρ 2 Ω; (12)
which is equivalent to
with the following boundary condition:
3
M.J. Mahmoodabadi Informatics in Medicine Unlocked 18 (2020) 100293
20 25
PSO-FVM PSO-FVM
HPM 20 HPM
15
OHPM OHPM
RKM 15 RKM
Susceptible Population ( )
Susceptible Population ( )
10
10
5
5
0
0
-5
-5
-10 -10
-15 -15
0 2 4 6 8 10 0 2 4 6 8 10
Time (T) Time (T)
(a) (b)
20 20
PSO-FVM PSO-FVM
HPM HPM
15 15
OHPM OHPM
RKM RKM
Susceptible Population ( )
Susceptible Population ( )
10
10
5
5
0
0
-5
-5
-10
-10
-15
0 2 4 6 8 10 0 2 4 6 8 10
Time (T) Time (T)
(c) (d)
Fig. 2. Susceptible population obtained by the PSO-FVM, HPM [23], OHPM and RKM [37] for (a) 5, (b) 6, (c) 9 and (d) 11 nodal points.
ψ ðε; λÞ ¼ ΛðεÞ Λðσ0 Þ þ λΛðσ0 Þ þ ΔðεÞ φðρÞ ¼ 0; (16) In the following, the implementation of the HPM to the nonlinear
differential equation of the epidemic model presented in Ref. [23] is
where, λ 2 ½0; 1� represents an embedding parameter, and σ 0 signifies an explained.
initial guess that meets the boundary conditions. It is obvious from the
ð1 λÞð_ε1 μ_ 0 Þ þ λð_ε1 þ βε1 ε2 Þ ¼ 0; (20)
above equations that
ψ ðε; 0Þ ¼ ΛðεÞ Λðσ0 Þ ¼ 0; ψ ðε; 1Þ ¼ ηðεÞ φðρÞ ¼ 0: (17) ð1 λÞð_ε2 ν_ 0 Þ þ λð_ε2 βε1 ε2 þ γ ε2 Þ ¼ 0; (21)
Moreover, the estimated solution of the governing equation is ε1;0 ðTÞ ¼ μ0 ; (23)
regarded as:
ε2;0 ðTÞ ¼ ν0 ; (24)
u ¼ limε ¼ ε0 þ ε1 þ ε2 þ …: (19)
λ→1
ε3;0 ðTÞ ¼ ζ0 ; (25)
4
M.J. Mahmoodabadi Informatics in Medicine Unlocked 18 (2020) 100293
45 45
PSO-FVM PSO-FVM
HPM HPM
40
40 OHPM OHPM
RKM RKM
Infectiive Population ( )
35
Infectiive Population ( )
35
30
30
25
25
20
20
15
15 10
0 2 4 6 8 10 0 2 4 6 8 10
Time (T) Time (T)
(a) (b)
45 45
PSO-FVM PSO-FVM
HPM 40 HPM
40
OHPM OHPM
RKM 35 RKM
35
Infectiive Population ( )
Infectiive Population ( )
30
30
25
25
20
20
15
15 10
10 5
0 2 4 6 8 10 0 2 4 6 8 10
Time (T) Time (T)
(c) (d)
Fig. 3. Infective population obtained by the PSO-FVM, HPM [23], OHPM and RKM [37] for (a) 5, (b) 6, (c) 9 and (d) 11 nodal points.
5
M.J. Mahmoodabadi Informatics in Medicine Unlocked 18 (2020) 100293
15 15
PSO-FVM PSO-FVM
14.5 HPM 14.5 HPM
OHPM OHPM
14 14
RKM RKM
Immune Population ( )
Immune Population ( )
13.5 13.5
13 13
12.5 12.5
12 12
11.5 11.5
11 11
10.5 10.5
10 10
0 2 4 6 8 10 0 2 4 6 8 10
Time (T) Time (T)
(a) (b)
15 15
PSO-FVM PSO-FVM
HPM 14.5 HPM
14 OHPM OHPM
14
RKM RKM
Immune Population ( )
Immune Population ( )
13.5
13
13
12 12.5
12
11
11.5
11
10
10.5
9 10
0 2 4 6 8 10 0 2 4 6 8 10
Time (T) Time (T)
(c) (d)
Fig. 4. Immune population obtained by the PSO-FVM, HPM [23], OHPM and RKM [37] for (a) 5, (b) 6, (c) 9 and (d) 11 nodal points.
Table 1 Table 3
Numerical values related to the susceptible population for ΔT ¼ 1. Numerical values related to the immune population for ΔT ¼ 1.
T PSO-FVM HPM OHPM RKM T PSO-FVM HPM OHPM RKM
1 17.693934 17.689495 17.699782 17.689502 ε_ 2;3 βν0 ε1;2 þ γε2;2 þ β2 μ0 ν0 T ε2;1 βμ0 ε2;2 ¼ 0; (37)
2 8.898461 20.246381 19.074785 20.246628
3 10.531942 22.525951 21.064423 22.526087 ε_ 3;1 γ ν0 ¼ 0; (38)
4 11.477740 24.444390 23.053494 24.431660
5 12.978053 26.017073 24.424092 25.924877
6 14.048613 27.396868 25.641427 27.016527 ε_ 3;2 γβμ0 ν0 T þ γ 2 ν0 T ¼ 0; (39)
7 15.292777 28.912437 26.338564 27.749515
8 16.257284 31.106532 26.814816 28.181573 1 1 2 2 1 3
9 17.202394 34.774300 27.102169 28.372475 ε_ 3;3 þ γβ2 μ0 ν20 T 2 γβ μ0 ν0 T 2 γ ν0 T 2 þ γ2 βμ0 ν0 T 2 ¼ 0: (40)
2 2 2
10 17.842218 41.001580 27.100427 28.376632
6
M.J. Mahmoodabadi Informatics in Medicine Unlocked 18 (2020) 100293
7
M.J. Mahmoodabadi Informatics in Medicine Unlocked 18 (2020) 100293
[26] Navaneeth B, Suchetha M. PSO optimized 1-D CNN-SVM architecture for real-time [32] Pal JK, Sankar Ray S, Pal SK. Fuzzy mutual information based grouping and new
detection and classification applications. Comput Biol Med 2019;108:85–92. fitness function for PSO in selection of miRNAs in cancer. Comput Biol Med 2017;
[27] Kartheeswaran S, Durairaj DDC. A data-parallelism approach for PSO-ANN based 89:540–8.
medical image reconstruction on a multi-core system. Inf Med Unlocked 2017;8: [33] Subasi A. Classification of EMG signals using PSO optimized SVM for diagnosis of
21–31. neuromuscular disorders. Comput Biol Med 2013;43:576–86.
[28] Abbasi E, Ghatee M, Shiri ME. FRAN and RBF-PSO as two components of a hyper [34] Biazar J. Solution of the epidemic model by Adomian decomposition method. Appl
framework to recognize protein folds. Comput Biol Med 2013;43:1182–91. Math Comput 2006;173(2):1101–6.
[29] Kumar Satapathy S, Dehuri S, lok Kumar Jagadev A. EEG signal classification using [35] Zhan Z, Zhang J, Li Y, Chung HS. Adaptive particle swarm optimization. IEEE
PSO trained RBF neural network for epilepsy identification. Inf Med Unlocked Trans Syst Man Cybern B 2009;39(6):1362–81.
2017;6:1–11. [36] He JH. Homotopy perturbation technique. Comput Methods Appl Mech Eng 1999;
[30] Gambhir S, Kumar Malik S, Kumar Y. PSO-ANN based diagnostic model for the 178:257–62.
early detection of dengue disease. New Horizons Transl Med 2017;4:1–8. [37] Devries PL, Hasbun JE. A first course in computational physics. second ed. Jones
[31] Joloudari JH, Saadatfar H, Dehzangi A, Shamshirband Sh. Computer-aided and Bartlett Publishers; 2011.
decision-making for predicting liver disease using PSO-based optimized SVM with
feature selection. Inf Med Unlocked 2019;17.