Nonlinear Dyn (2020) 101:1667–1680

https://doi.org/10.1007/s11071-020-05743-y (0123456789().,-volV)
( 01234567
89().,-volV)

ORIGINAL PAPER

SEIR modeling of the COVID-19 and its dynamics

Shaobo He . Yuexi Peng . Kehui Sun

Received: 25 April 2020 / Accepted: 4 June 2020 / Published online: 18 June 2020
Ó Springer Nature B.V. 2020

Abstract In this paper, a SEIR epidemic model for 1 Introduction

the COVID-19 is built according to some general
control strategies, such as hospital, quarantine and At the end of 2019, a novel coronavirus disease
external input. Based on the data of Hubei province, (COVID-19) was declared as a major health hazard by
the particle swarm optimization (PSO) algorithm is World Health Organization (WHO). At present, this
applied to estimate the parameters of the system. We disease is rapidly growing in many countries, and the
found that the parameters of the proposed SEIR model global number of COVID-19 cases is increasing at a
are different for different scenarios. Then, the model is rapid rate. This coronavirus is a kind of enveloped,
employed to show the evolution of the epidemic in single stranded and positive sense virus which belongs
Hubei province, which shows that it can be used to to the RNA coronaviridae family [1, 2]. In early
forecast COVID-19 epidemic situation. Moreover, by December of 2019, this infectious disease has begun to
introducing the seasonality and stochastic infection outbreak in Wuhan, the capital city of Hubei province,
the parameters, nonlinear dynamics including chaos China. Until now, the epidemic in China is basically
are found in the system. Finally, we discussed the under control, but there are still many infections
control strategies of the COVID-19 based on the around the world. To defeat the epidemic, scientists in
structure and parameters of the proposed model. different fields investigated the COVID-19 from
different points of view. Those aspects include
Keywords COVID-19  Coronavirus  SEIR model  pathology, sociology perspective, the infection mech-
Nonlinear dynamics  Control anism and prediction [3–8].
In the history of mankind, there are many other
outbreak and transmission of diseases such as dengue
fever, malaria, influenza, pestilence and HIV/AIDS.
How to built a proper epidemiological model for these
epidemics is a challenging task. Some scientists treat
the disease spread as a complex network for forecast
and modeling [9, 10]. For the COVID-19, Bastian
Prasse et al. [10] designed a network-based model
which is built by the cities and traffic flow to describe
S. He  Y. Peng (&)  K. Sun
School of Physics and Electronics, Central South
the epidemic in the Hubei province. At present, the SIS
University, Changsha 410083, China [11, 12], SIR[13] and SEIR [14, 15] models provide
e-mail: pyx244896301@163.com

123
1668 S. He et al.

another way for the simulation of epidemics. Lots of 2 SEIR modeling of the COVID-19
research works have been reported. It shows that those
SIS, SIR and SEIR models can reflect the dynamics of The classical SEIR model has four elements which are
different epidemics well. Meanwhile, these models S (susceptible), E (exposed), I (infectious) and R
have been used to model the COVID-19 [16, 17]. For (recovered). Thus, N ¼ S þ E þ I þ R means the total
instance, Tang et al. [17] investigated a general SEIR- number of people. The basic hypothesis of the SEIR
type epidemiological model where quarantine, isola- model is that all the individuals in the model will have
tion and treatment are considered. Moreover, there are the four roles as time goes on. The SEIR model has
also other methods for modeling of the COVID-19 some limitations for the real situations, but it provides
[18, 19]. Wang et al. [19] applied the phase-adjusted a basic model for the research of different kinds of
estimation for the number of coronavirus disease 2019 epidemic.
cases in Wuhan. Thus in this paper, we try to propose a Based on the basic SEIR model, we proposed a new
SEIR model to simulate the process of COVID-19. model which is denoted by
Chaos widely exits in nature and man-made 8
> S
systems including those biological systems [20–24]. >
> S_ ¼  ðb1 I1 þ b2 I2 þ vEÞ þ q1 Q  q2 S þ aR
>
> N
According to the famous Logistic map, it shows that >
>
>
> S
the natural evolution of the population size could be >
>
> E_ ¼ ðb1 I1 þ b2 I2 þ vEÞ  h1 E  h2 E
>
> N
chaotic. However, it is not a good thing to find chaos in < _
I 1 ¼ h1 E  c 1 I 1 ;
the SEIR model. Unfortunately, chaos in the SIR, SIS >
and SEIR models has been investigated by many > I_2 ¼ h2 E  c2 I2  uI2 þ kðK þ QÞ
>
>
>
>
researchers. Generally, the seasonality and stochastic >
>
> R_ ¼ c1 I1 þ c2 I2 þ /H  aR
>
>
infection are introduced to the system for the nonlinear >
> H_ ¼ uI2  /H
>
:
dynamics. For instance, Kuznetsov and Piccardi [25] Q_ ¼ K þ q2 S  kðK þ QÞ  q1 Q
investigated the bifurcations of the periodic solutions
ð1Þ
of SEIR and SIR epidemic models with sinusoidally
varying contact rate. Meanwhile, the fractional-order where S, E, I1 , I2 , R, H and Q are the system variables.
SEIR epidemic model has aroused research interests The descriptions of those variables are presented in
of scientists. He et al. [26] investigated the epidemic Table 1. The description of the system parameters is
outbreaks using the SIR model, and a hard limited illustrated in Table 2, and the relationship between
controller is designed for the control of the system. different variables is shown in Fig. 1. In this model,
However, on the one hand, those SIR and SEIR the infectious class is divided into two parts, I1 and I2 .
models cannot always show the nature of the COVID- Meanwhile, we consider the quarantined class and
19, and we need to modify the system. On the other hospitalized class in the model according to the real
hand, the nonlinear dynamics of the system should be situation. For example, if one got coronavirus-like
investigated. Thus, we need to get more information
on the dynamics of the epidemic system.
The rest of this paper is organized as follows. In
Sect. 2, the modified SEIR model is designed and the Table 1 Description of the system variables
descriptions of the system are presented. In Sect. 3, the Variable Description
SEIR model is applied to the COVID-19 data of Hubei
province where the PSO algorithm is introduced to S Susceptible class
estimate the parameters. In Sect. 4, the seasonality and E Exposed
stochastic infection are introduced to the model and I1 Infectious without intervention
the dynamics of the system is investigated. In Sect. 5, I2 Infectious with intervention
the structure, parameters on the dynamics of the R Recovered
system and how to control the epidemic of the system Q Quarantined
are discussed. Section 6 is the summary of the H Hospitalized
analysis.

123
SEIR modeling of the COVID-19 and its dynamics 1669

Table 2 Description of the system parameters

Parameters Description

a Temporary immunity rate

b1 , b2 The contact and infection rate of transmission per contact from infected class
v Probability of transmission per contact from exposed individuals
h1 , h2 Transition rate of exposed individuals to the infected class
c1 , c2 Recovery rate of symptomatic infected individuals to recovered
u Rate of infectious with symptoms to hospitalized
/ Recovered rate of quarantined infected individuals
k Rate of the quarantined class to the recovered class
q1 , q2 Transition rate of quarantined exposed between the quarantined infected class and the wider community
K External input from the foreign countries

K. Obviously, N is not a constant and it varies over

time when K 6¼ 0. Since there are a large population
under voluntary home quarantine. Thus, for a chosen
place, N is not the total population of that place, but it
can be estimated by adding the final number of
recoveries and deaths.

3 Estimation of the model parameters

Fig. 1 Flowchart of the proposed SEIR model for COVID-19 3.1 The PSO algorithm

symptoms or comes from other place like abroad, he or Particle swarm optimization (PSO) algorithm is a
she needs to be in quarantine for at least 14 days. famous population-based stochastic optimization
Obviously, as shown in Fig. 1, we consider two algorithm motivated by intelligent collective behav-
main channels in the proposed model. The first one ior, such as the foraging process of bird group [27]. In
goes to S ! E ! I1 ! R, and the second channel PSO algorithm, each particle represents a bird, and the
goes to S ! Q ! I2 ! H ! R. The first case shows algorithm starts with a random initialization of the
the natural process of the epidemic, and it is a typical particle locations. For one iteration, each particle
SEIR model. The second channel considers the control keeps track of its own best position and the popula-
from the government including the quarantine and tion’s best position to update its position and velocity.
hospital. As a result, the designed model is an Considering a one-dimensional optimization problem,
improved version of the SEIR model. the velocity and the position of the particle i is defined
If there is no quarantine (q2 ¼ 0), hospital treat- by
ment / ¼ 0 and the recovered is immune to the virus
(a ¼ 0), the model becomes to the classical SEIR Vi ðt þ 1Þ ¼ xðtÞVi ðtÞ þ c1 r1 ½Pb;i ðtÞ  Xi ðtÞ
; ð2Þ
model. However, there are always quarantine and þc2 r2 ½Pg ðtÞ  Xi ðtÞ
hospital treatment. Meanwhile, there is no evidence
and
that the recovered is immune to the COVID-19. Thus,
we need to considered these factors in the model. In X i ðt þ 1Þ ¼ X i ðtÞ þ V i ðt þ 1Þ; ð3Þ
this paper, we have N 6¼ S þ E þ I1 þ I2 þ R þ Q þ
H according to Eq. (1) since there is an external input where Vi ðtÞ and Xi ðtÞ are velocity and position of the
particle i at the t-th iteration, respectively. c1 and c2 are

123
1670 S. He et al.

the learning factors. r1 and r2 are random numbers Step 7: Repeat the Steps 3–6 until the termination
between 0 and 1. Pb;i ðtÞ represents the best position of criterion is satisfied.
the particle i at the t-th iteration, and Pg ðtÞ represents
the population’s best position at the t-th iteration. x is
3.2 Parameter estimation
called inertia weight, which is very important for the
search process of PSO algorithm. Here, it is defined by
In this section, through the actual COVID-19 data
[28]
from Hubei province, the PSO algorithm is utilized to
2 estimate the parameters of the proposed SEIR model
xðtÞ ¼ ðxmax  xmin Þe500ðt=TÞ þ xmin ; ð4Þ
to fit the real situation. The COVID-19 data come from
where t and T are the current and maximum iterations the official website of the Wuhan Municipal Health
of the PSO algorithm, respectively, and xmin and xmax Commission (http://wjw.wh.gov.cn/), and some actual
are the minimal and maximum value of the inertia data are listed in Table 3.
weight, respectively. Suppose that we meet a mini- In the face of the pressure of epidemic prevention
mum optimization problem, the implementation steps and control, Wuhan government announced to seal off
of the PSO algorithm are summarized as follows: the city from all outside contact on January 23rd, 2020.
Then, other cities in Hubei province also took the
Step 1: The position of each particle is randomly
‘‘closure city’’ measure. The COVID-19 epidemic
initialized.
situation of Hubei is relatively stable after January
Step 2: Calculate the fitness value FðXi ðtÞÞ of the
23rd, 2020, so we chose to study the data between
particle i, and find the Pb;i ðtÞ and the Pg ðtÞ.
January 24th and April 12th.
Step 3: If FðXi ðtÞÞ\FðPb;i ðtÞÞ, then replace the
The initial values setting of SEIR model is
Pb;i ðtÞ by the Xi ðtÞ.
presented in Table 4, where N is the total population
Step 4: If FðXi ðtÞÞ\FðPg ðtÞÞ, then replace the Pg ðtÞ
of Hubei affected by the COVID-19 epidemic in
by the Xi ðtÞ. January 24th, 2020, and E is calculated based on the
Step 5: Calculate the inertia weight by Eq. (4). number of confirmed patients. I1 is an estimated value
Step 6: Update velocity and position of the particle
i according to Eqs. (2) and (3), respectively.

Table 3 Actual COVID-19 data from Hubei (January 24th to February 8th)
Date Cumulative infected cases Cumulative deaths Cumulative recovered cases Current quarantined

2020/1/24 729 39 32 4711

2020/1/25 1052 52 42 6904
2020/1/26 1423 76 44 9103
2020/1/27 2714 100 47 15,559
2020/1/28 3554 125 80 20,366
2020/1/29 4586 162 90 26,632
2020/1/30 5806 204 116 32,340
2020/1/31 7153 249 166 36,838
2020/2/1 9074 294 215 43,121
2020/2/2 11,177 350 295 48,171
2020/2/3 13,522 414 396 58,544
2020/2/4 16,678 479 520 66,764
2020/2/5 19,665 549 633 64,127
2020/2/6 22,112 618 817 64,057
2020/2/7 24,953 699 1115 67,802
2020/2/8 27,100 780 1439 70,438
......

123
SEIR modeling of the COVID-19 and its dynamics 1671

Table 4 Initial values of the SEIR model

Parameter N E I1 I2 H R Q K

Value 6:5563  104 5077 I2  0:01 729 658 32 4711 10

Table 5 System Parameter Values 5400

parameters of SEIR model
b1 1:0538  101 5350
b2 1:0538  101
v 1:6221  101 5300

Error
q1 3
2:8133  10
5250
q2 1:2668  101
h1 9:5000  104 5200
h2 3:5412  102
c1 8:5000  103 5150
0 20 40 60 80 100
c2 3
1:0037  10 Iteration
k 9:4522  102
Fig. 2 The error convergence curve of PSO algorithm in the
a 1:2048  104 outbreak stage

based on I2 , and the other initial values are originated × 104

2.6
from the actual data.
2.4
The system parameters of SEIR model are calcu-
lated by the actual data, as shown in Table 5. 2.2
However, there is no accurate statistics of the rate of 2
Error

infectious to hospitalized u and the recovered rate of 1.8

quarantined infected individuals /. Here, the two
1.6
parameters are estimated by the PSO algorithm with
the actual data of R and H. The settings of PSO 1.4

algorithm are set as: population of particle swarm 1.2

0 20 40 60 80 100
NP ¼ 40, learning factors c1 ¼ c2 ¼ 2, maximal iter- Iteration
ation T ¼ 100 and the search spaces u; / 2 ð0; 0:1.
The COVID-19 epidemic situation in Hubei is Fig. 3 The error convergence curve of PSO algorithm in the
inhibition stage
divided into two stages: the outbreak stage (the first 19
days) and the inhibition stage (the 20th day to the end).
there are some errors between the estimated number
In the outbreak stage, according to the actual data of
and the actual number, it shows that the estimated
R and H, the u and the / are estimated to u ¼ 0:2910,
values match well with the real situation. However, the
/ ¼ 0:0107, respectively. The error convergence
accuracy is not satisfying in the second stage which
curve of PSO is shown in Fig. 2. After the outbreak
shows that the real data are smaller than the estimated
stage, due to the continuous assistance from other
values, but the trend is basically the same.
provinces and other countries, the epidemic in Hubei
There are two reasons for the deviation. One is that
began to enter the inhibition stage. In this stage, the
only two parameters, namely the rate of infectious to
error convergence curve of PSO is given in Fig. 3, and
hospitalized u and the recovered rate of quarantined
the estimated u and / changed to
infected individuals /, are estimated, while the rest of
u ¼ 0:0973; / ¼ 0:0416, respectively.
the parameters are set as a matter of experience.
The estimated and actual trajectories in the two
Moreover, the control measures for containing the
stages are shown in Fig. 4. In the first stage, although

123
1672 S. He et al.

Fig. 4 Actual and estimated

trajectories for the epidemic
situation in Hubei province

outbreak are more and more powerful; thus, the system The seasonality is widely found in the epidemic
parameter should be time varying variables. For models [26, 29], and it can make the system more
instance, compared with the outbreak stage, the complex. Indeed, there is no report showing that the
hospitalization rate decreased a lot, and the cure rate effect of seasonality for COVID-19 spread since this
increased nearly four times, which means that the epidemic outbreaks only about half year until now.
number of confirmed infected cases is declining a lot But we try to introduce the seasonality to the system
and the number of patients recovering is increasing and analyze chaos in the system from an academic
rapidly. point of view. Meanwhile, there are individual differ-
ences and many unpredictable factors in the epidemic
infection. Thus, the noise is an important factor
4 Nonlinear dynamics of the model considered in our analysis. Here, three cases are
analyzed to show how the system parameter a,
4.1 SEIR model with seasonality and stochastic seasonality and stochastic infection affect the dynam-
infection ics of the system.
Case 1: The parameter b1 contains seasonality and
The 0–1 test algorithm is employed to verify the stochastic infection, and the three contact and infec-
existence of chaos in the model. If a set of discrete time tion rate parameters are defined as
series xðnÞðn ¼ 1; 2; 3; . . .Þ represents a one-dimen- 
b2 ¼ 30:03; v ¼ 30:40
sional observable data set obtained from the modified ; ð6Þ
SEIR system, then the following two real-valued b1 ðtÞ ¼ b0 ð1 þ e1 sinð2ptÞ þ e2 nðtÞÞ
sequences are defined as [30] where b0 ¼ 2  b1 ¼ 60, e1 and e2 are degree of the
( P seasonality and stochastic infection, respectively.
pðnÞ ¼ nj¼1 xð jÞ cosðhð jÞÞ
P ; ð5Þ hnðtÞi is the white Gaussian independent noises, and
sðnÞ ¼ nj¼1 xð jÞ sinðhð jÞÞ it has the properties of hnðtÞi ¼ 0 and
P j   hnðtÞ; nðsÞi ¼ dðt  sÞ.
where hð jÞ ¼ jg þ xðiÞ, and g 2 p5 ; 4p 5 . By plot-
i¼1 The analysis results of the system with different
ting the trajectories in the (p, s)-plane, the state of the parameters are shown in Fig. 5. In Fig. 5a, b, we set
system can be identified. Usually, the bounded trajec- e1 ¼ 0, e2 ¼ 0 and a ¼ 0:08. It shows that the system
tories in the (p, s)-plane imply the dynamics of the is convergent to a limited region. Thus, the system is
time series is regular, while Brownian-like (un- not chaotic without the seasonality and stochastic
bounded) trajectories imply chaos. infection. When e1 ¼ 0:8 and e2 ¼ 0, the system has
seasonality but no stochastic infection. It shows that

123
SEIR modeling of the COVID-19 and its dynamics 1673

(a) (b)

(c) (d) (e)

(f) (g) (h)

Fig. 5 Dynamics analysis results of case 1 with different enlarged drawing d and p  s plot e with e1 ¼ 0:8, e2 ¼ 0 and
parameters. Phase diagram a and time series b of the system a ¼ 0:08; phase diagram f, its partial enlarged drawing g and
with e1 ¼ 0, e2 ¼ 0 and a ¼ 0:08; phase diagram c, its partial p  s plot h with e1 ¼ 0:8, e2 ¼ 0:2 and a ¼ 0:08

the system now is chaotic according to the p  s plot. where b0 ¼ 2  b2 ¼ 60, e1 and e2 are the degree of
As shown in Fig. 5c that the attractor looks like a the seasonality and stochastic infection, respectively.
limited circle. However, there are many lines in the Firstly, we let e1 ¼ 0:8 and e2 ¼ 0. If a ¼ 0:02,
‘‘circle’’ as it is illustrated in Fig. 5d. If e1 ¼ 0:8 and a ¼ 0:03, a ¼ 0:04747 and a ¼ 0:08, different kinds
e2 ¼ 0:2, there are both seasonality and stochastic of chaotic attractors are shown in Fig. 6a–d, respec-
infection in the system. At this time, the complex tively. Meanwhile, we let e1 ¼ 0:8 and e2 ¼ 0:2, the
dynamics is observed in the system, as shown in chaotic attractors with different values of a are shown
Fig. 5f–h. in Fig. 6e–h. The p  s plots of attractors of Fig. 6b–d
Case 2: The parameter b2 contains seasonality and are shown in Fig. 6i–k, respectively. It verifies the
stochastic infection, while the other two infection rate existence of chaos in the model.
parameters are constants. Thus, the parameters are Case 3: The parameter v1 contains seasonality and
defined as stochastic infection, and b1 and b2 are contacts. As a

b1 ¼ 30; v ¼ 30:40 results, the parameters in this case are given by
; ð7Þ
b2 ðtÞ ¼ b0 ð1 þ e1 sinð2ptÞ þ e2 nðtÞÞ

123
1674 S. He et al.

(a) (b) (c) (d)

(e) (f) (g) (h)

(i) (j) (k)

Fig. 6 Dynamics analysis results of case 2. Phase diagrams a ¼ 0:02 e, a ¼ 0:03 f, a ¼ 0:04747 g and a ¼ 0:08 h; p  s
with e1 ¼ 0:8, e2 ¼ 0 and a ¼ 0:02 a, a ¼ 0:03 b, a ¼ 0:04747 c plots of e1 ¼ 0:8, e2 ¼ 0 and a ¼ 0:03 i, a ¼ 0:04747 j and
and a ¼ 0:08 d; phase diagrams with e1 ¼ 0:8, e2 ¼ 0:2 and a ¼ 0:08 k


b1 ¼ 30; b2 ðtÞ ¼ 30 To investigate chaos in the proposed model, the
ð8Þ parameters are set as b1 ¼ 30, b2 ¼ 30:0300,
v ¼ v0 ð1 þ e1 sinð2ptÞ þ e2 nðtÞÞ
v ¼ 30:40, q1 ¼ 1=14, q2 ¼ 0:002, h1 ¼ 20:054,
where b0 ¼ 2v ¼ 60:8, e1 and e2 are the degree of the h2 ¼ 20:12, u ¼ 0:00009, / ¼ 0:8, k ¼ 0:4 and
seasonality and stochastic infection, respectively. Let N ¼ 106 . The initial conditions of the model are given
e1 ¼ 0:8, e2 ¼ 0 and a ¼ 0:0133, the phase diagram is by ½S; E; I1 ; I2 ; R; H; Q ¼ ½94;076; 4007; 262; 524; 31;
shown in Fig. 7a. It shows in Fig. 7b that the attractor 100; 1000.
is chaotic. If we set e2 ¼ 0:2, the phase diagram is
illustrated in Fig. 7c, where a much wider range in the 4.2 Bifurcation analysis of case 2
phase space when the system has stochastic infection.
When e1 ¼ 0:8, e2 ¼ 0 and a ¼ 0:08, the phase To further analyze dynamics of the system, we choose
diagram is shown in Fig. 7d, while the time series case 2 as an example to show the bifurcations of the
are shown in Fig. 7e. proposed system. The parameters are set as b1 ¼ 30,
The proposed system is an improved SEIR system b2 ¼ 30:0300, v ¼ 30:40, q1 ¼ 1=14, q2 ¼ 0:002,
with quarantined class and hospitalized class. As with h1 ¼ 20:054, h2 ¼ 20:12, u ¼ 0:00009, / ¼ 0:8, k ¼
other SIR and SEIR models, this model can also 0:4 and N ¼ 106 . The initial conditions of the model
generate chaos with given parameters. And the are given by ½S; E; I1 ; I2 ; R; H; Q=[94076, 4007, 262,
existence chaos is verified by the 0–1 test method. 524,31, 100, 1000].
In this section, we consider the dynamics of the Firstly, let e1 ¼ 0:8, e2 ¼ 0, and the parameter a
proposed SEIR model. In Refs. [26, 29], the param- varies from 0.02 to 0.08 with step size of 0.00024. The
eters of the SEIR model are set where the infection rate bifurcation diagram is shown in Fig. 8a. Meanwhile, if
b is set as quite large values. For instance, in Ref. [29], e2 ¼ 0:2, the corresponding bifurcation diagram is
b ¼ 108 for the proposed SEIR Dengue fever model.

123
SEIR modeling of the COVID-19 and its dynamics 1675

(a) (b) (c)

(d) (e)

Fig. 7 Dynamics analysis results of case 3. Phase diagram a and p  s plot b with e1 ¼ 0:8, e2 ¼ 0 and a ¼ 0:0133; phase diagram c
with e1 ¼ 0:8, e2 ¼ 0:2 and a ¼ 0:0133; phase diagram d and time series e with e1 ¼ 0:8, e2 ¼ 0:2 and a ¼ 0:08

shown in Fig. 8b. Obviously, the bifurcation diagram For a given time series {x(n), n=0, 1, 2,   , L-1}
of Fig. 8a has no noise, while that in Fig. 8b does. It with a length of L, let xðnÞ ¼ xðnÞ  x, where x is the
shows that the stochastic infection makes the system mean value of time series. Its corresponding discrete
fluctuate more volatile. Fourier transform (DFT) is defined by
Secondly, let a ¼ 0:08, e2 ¼ 0, and the parameter
X
L1
e1 ¼ 0:8 varies from 0.02 to 0.08 with step size of XðkÞ ¼ xðnÞej2pnk=L ; ð9Þ
0.004. The bifurcation diagrams with e1 are shown in n¼0
Fig. 9. It shows that when there exists stochastic
where k ¼ 0; 1;    ; L  1 and j is the imaginary unit.
infection, the bifurcation diagram shows more com-
If the power of a discrete power spectrum with the kth
plex behaviors of the system.
As shown in Figs. 8 and 9, it shows that the system frequency is jXðkÞj2 , then the ‘‘probability’’ of this
has rich dynamics with both parameters a and e1 . frequency is defined as
When the system has stochastic infection, the system
jXðkÞj2
will become more complex. We hold the opinion that Pk ¼ PL=21 2
: ð10Þ
chaos is also the nature of the system, and the k¼0 jXðkÞj
seasonality and the temporary immunity rate can When the DFT is employed, the summation runs from
change the dynamics of the system. k ¼ 0 to k ¼ N=2  1. The normalization entropy is
denoted by [31]
4.3 Complexity of the case 2
  1 X
N=21

In this section, the spectral entropy (SE) algorithm SE xL ¼ Pk lnðPk Þ; ð11Þ

lnðN=2Þ k¼0
[31] is employed to analyze complexity of the
proposed SEIR system, and steps are presented as where lnðN=2Þ is the entropy of a completely random
follows. signal. Obviously, the more balanced the probability

123
1676 S. He et al.

105
4
105 size of 0.0036. The complexity analysis results with e1
(a) (b)
3.8
are shown in Fig. 10c, d. Here, e2 ¼ 0 is employed in
3.5
3.6 Fig. 10c, while e2 ¼ 0:2 is used in Fig. 10d. The

Smax
Smax

3.4
3 complexity analysis results with parameters show that
the stochastic infection does not affect the complexity
3.2 2.5
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.02 0.03 0.04 0.05 0.06 0.07 0.08 of the system. The system has higher complexity when
a takes values between 0.03 and 0.06, and e1 takes
Fig. 8 Bifurcation diagrams of the system with the variation of those values which are larger than 0.6.
parameter a a e1 ¼ 0:8, e2 ¼ 0; b e1 ¼ 0:8, e2 ¼ 0:2 Fix e2 ¼ 0, vary a from 0.02 to 0.08 with step size
of 0.0006 and e1 varies from 0.1 to 1 with step size of
0.009. The complexity analysis result in the a  e1
plane is shown in Fig. 11. Obviously, the higher
complexity region is located in the right side of the
parameter plane, where a 2 ½0:4; 1 and
(a) (b) e1 2 ½0:025; 0:055.
Since the complexity measure results are obtained
based on the generated time series, MSE provides an
Fig. 9 Bifurcation diagrams of the system with the variation of effective way for the dynamics analysis of the system.
parameter e1 a a ¼ 0:08, e2 ¼ 0; b a ¼ 0:8, e2 ¼ 0:2 When there is higher complexity, the behavior of the
model is more complex, vice versa. For a epidemic
distribution is, the higher complexity (the larger system, high complexity means outbreak. Thus, we
entropy) the time series is. The larger measuring value can use the complexity measure algorithm to monitor
means higher complexity, and vice versa. the dynamics of the proposed SEIR system.
Based on the above complexity algorithms, multi-
scale complexity algorithm is designed. For a one-
dimensional discrete time series 5 Discussion
fxðnÞ : n ¼ 0; 1; . . .; N  1g, the consecutive coarse-
grained time series are constructed by [32] In this paper, the parameters of the system are mainly
chosen by two means including the references and the
1 X
js1
ys ðjÞ ¼ xðjÞ; ð12Þ PSO algorithm. Usually, the parameters can be set
s ðj1Þs according to the existing work. For instance, the
contact and infection rate parameters are defined
where 1  j  ½N=s, s is the scale factor which
according to Refs. [26, 29]. Also, there are some other
represents the length of the non-overlapping window,
references which show some parameters of the system.
and ½ denotes the floor function. Obviously, when
Moreover, since there is actual COVID-19 data from
s ¼ 1, the sequence ys is the original sequence
Hubei province, we use the PSO algorithm to estimate
fxðnÞ; n ¼ 0; 2; 3;    ; L  1g. Thus, the complexity
the parameters of the system. As shown above, the
of y1 is the complexity of the original sequence. In this system has rich dynamics with the given parameters,
paper, the multiscale complexity is defined as [31] especially when the parameters b1 , b2 and v have
1 X
smax seasonality and stochastic infection. The existence of
MSE ¼ SEðys Þ: ð13Þ chaos is verified by the 0–1 test, and complexity of the
smax s¼1
generated time series is measured. Here, different sets
In this paper, we set smax ¼ 20. of parameters are summarized in Table 6. It should
Fix e1 ¼ 0:8 and vary the parameter a from 0.02 to noted that, when the parameters are set as the set A,
0.08 with step size of 0.00024. MSE complexity they seem ‘‘large.’’ However, those parameters should
analysis results are shown in Fig. 10a, b, where be multiplied by S/N. Thus, they are reasonable for the
Fig. 10a for e2 ¼ 0 and Fig. 10b for e2 ¼ 0:2. Fix a ¼ proposed model. According to our analysis, it shows
0:08 and the parameter e1 varies from 0.1 to 1 with step that chaos are found in the system with those ‘‘large
parameters.’’ In fact, we want to explore the nonlinear

123
SEIR modeling of the COVID-19 and its dynamics 1677

(a) (b)

(c) (d)

Fig. 10 MSE analysis results of the system with the variation of parameters a and e1 a e1 ¼ 0:8, e2 ¼ 0 and a varying; b e1 ¼ 0:8,
e2 ¼ 0:2 and a varying; c a ¼ 0:08, e2 ¼ 0 and e1 varying; d a ¼ 0:08, e2 ¼ 0:2 and e1 varying

magenta color lines (M), blue color lines (B), red color
lines (R) and green color lines (G) are obtained using
the following parameters:
M k ¼ 0:0004, u ¼ 0:009, a ¼ 0:0,
B k ¼ 0:04, u ¼ 0:009, a ¼ 0:0,
R k ¼ 0:0004, u ¼ 0:8, a ¼ 0:0,
G k ¼ 0:04, u ¼ 0:8, a ¼ 0:5.

It shows that the number of infected class (I1 , I2 ) and

hospitalized class (H) is different with different
parameters. When u ¼ 0:0009, there is a peak value
for I2 . It means that if the hospital reception capacity is
limited, the infected class (I2 ) will increase dramati-
Fig. 11 MSE analysis results of the system with the variation of
cally. However, when u ¼ 0:8, the infected class (I2 )
both parameters a and e1 keeps a relative low level; thus, the infection can be
controlled well. As shown in Fig. 12, when a ¼ 0:5, it
dynamics of the proposed system and to study how is quite difficulty for the system to become conver-
does chaos occur in the system. gent. The reason is obvious because those recovered
Figure 12 shows the evolution of the system with can be infected again, and a closed loop system is
different parameters. The parameters used the Set D, observed. Because no reports show that the recovered
and different colors lines in the figure including class is immune to the COVID-19, we need to be

123
1678 S. He et al.

Table 6 Values of the Parameters Set A (Chaos) Set B (Stage 1) Set C (Stage 2) Set D (Test)
parameters for different
cases a 0.08 1:2048  104 1:2048  104 0 or 0.5
b1 , 30 1:0538  101 1:0538  101 0.01
b2 30.03 1 1 0.3
1:0538  10 1:0538  10
v 30.40 1:6221  101 1:6221  101 0.4
h1 , 20.054 9:5000  104 9:5000  104 0.01
h2 20.12 2 2 0.02
3:5412  10 3:5412  10
c1 , 26, 8:5000  103 8:5000  103 5  102
c2 26, 1:0037  103 1:0037  103 6  102
u 0.00009 0.2910 0.0973 0.009
/ 0.8 0.0107 0.0416 0.008
k 0.4 9:4522  102 9:4522  102 4  104
q1 , 1/14, 3 3 1/14
2:8133  10 2:8133  10
q2 0.002, 1 1 0.002
1:2668  10 1:2668  10
K 10 10 10 10 or 100

(a) (b) (c)

(d) (e) (f)

Fig. 12 Evolution of the system with different parameters. a I1 with K ¼ 10; b I2 with K ¼ 10; c H with K ¼ 10; d I1 with K ¼ 100; e
I2 with K ¼ 100; f H with K ¼ 100

aware of those recovered to be infected again. Here, makes the system hard to converge. Thus, it shows in
the evolution of the system with parameters of set E is Fig. 13 that these variables converge to zero slowly.
shown in Fig. 13, which shows how all the classes of In the early stage of the COVID-19 epidemic, the
the system affect the dynamics. Generally, all the epidemic situation in Hubei province presents an
classes except the recovered class R will converge to uncontrollable trend. However, due to the low popu-
zero. However, Fig. 13 is simulated with external lation contact rate, high hospitalization rate and high
input K ¼ 100 and a ¼ 0:5. Since there is no evidence cure rate, the epidemic was quickly controlled after 20
that the recovered class is immune to the virus, this days. Therefore, the government’s attention, people’s

123
SEIR modeling of the COVID-19 and its dynamics 1679

(1) The proposed model has considered the quar-

antine and treatment, so it is more suitable for
the dynamics of the epidemic of COVID-19.
(2) The PSO algorithm provides a good way for
parameter estimation of the SEIR model. And
according to the application to the data of Hubei
province, the accuracy is acceptable. The main
trends of the epidemic evolution are illustrated.
(3) Nonlinear dynamics of the system is investi-
gated by means of bifurcation diagram, MSE
algorithm and 0–1 test algorithm. It shows that,
for the given parameters, if there exists season-
ality and stochastic infection, the system can
generate chaos.
(4) Some control suggestions are suggested based
Fig. 13 Evolution of the system with parameters of set E on the proposed model. Meanwhile, we found
that the dynamics of the system is different with
self-awareness and sufficient medical resources are the different sets of parameters.
key to eliminate the threat of COVID-19.
To get better estimation results, we need to built a Acknowledgements This work was supported by the Natural
proper model and also need to set proper parameters Science Foundation of China (Nos. 61901530, 11747150), the
for the systems. To the knowledge of authors, the China Postdoctoral Science Foundation (No.2019M652791)
parameters of the system change as time since the and the Postdoctoral Innovative Talents Support Program (No.
BX20180386). The authors would like to thank the editor and
control from the government is different along with the referees for their carefully reading of this manuscript and for
time. Thus, we can also treat the parameters as their valuable suggestions.
functions of time. If the values of b1 , b2 and v are
large, the system can even become chaotic. When q2 Compliance with ethical standards
takes larger values, it means that there are more people
Conflict of interest The authors declare that they have no
which have like COVID-19-symptoms. In fact, the conflict of interest.
government should take stronger and harsher mea-
sures to increase isolation, especially there are many
potential infections. Meanwhile, if the quarantine is References
done well, the values of b1 , b2 and v will be also much
smaller; thus, it is helpful to control the spread of the 1. Chen, Y., Liu, Q., Guo, D.: Emerging coronaviruses: gen-
epidemic disease. ome structure, replication, and pathogenesis. J. Med. Virol.
92, 418–423 (2020)
2. Spaan, W.J.M., Cavanagh, D., Horzinek, M.C.: Coron-
aviruses: structure and genome expression. J. General Virol.
6 Conclusion 69, 2939–52 (1988)
3. Zhu, N., Zhang, D., Wang, W., et al.: A novel coronavirus
from patients with pneumonia in china, 2019. New Engl.
In this paper, a SEIR model is proposed for the J. Med. 382, 727–733 (2020)
COVID-19. Parameters of the system are estimated by 4. Shimizu, K.: 2019-ncov, fake news, and racism. Lancet 395,
the PSO algorithm, and dynamics of the system is 685–686 (2020)
investigated. Finally, how the parameters affect the 5. Kock, R.A., Karesh, W., Veas, F., Velavan, T.P., Simons,
D., Mboera, L.E.G., Dar, O.A., Arruda, L.B., Zumla, A.:
dynamics of the system is discussed and the control 2019-ncov in context: lessons learned? Lancet. Planet.
strategies are presented. The conclusions of this paper Health 4, 87–88 (2020)
are given as follows. 6. Ceraolo, C., Giorgi, F.M.: Genomic variance of the
2019-ncov coronavirus. J. Med. Virol. 92, 522–528 (2020)

123
1680 S. He et al.

7. Ji, W., Wang, W., Zhao, X., et al.: Cross-species transmis- 22. Aihara, K., Suzuki, H.: Theory of hybrid dynamical systems
sion of the newly identified coronavirus 2019-ncov. J. Med. and its applications to biological and medical systems.
Virol. 92, 433–440 (2020) Philosoph. Trans. R. Soc. A Math. Phys. Eng. Sci. 368,
8. Fanelli, D., Piazza, F.: Analysis and forecast of covid-19 4893–4914 (2010)
spreading in china, italy and france. Chaos Solitons Fractals 23. Zhang, Q., Liu, C., Zhang, X.: Complexity, analysis and
134, 109761–109761 (2020) control of singular biological systems (2012)
9. Keeling, M.J., Eames, K.T.D.: Networks and epidemic 24. He, S., Fataf, N.A.A., Banerjee, S., et al.: Complexity in the
models. J. R. Soc. Interface 2, 295–307 (2005) muscular blood vessel model with variable fractional
10. Prasse, B., Achterberg, M.A., Ma, L. et al.: Network-based derivative and external disturbances. Physica A 526,
prediction of the 2019-ncov epidemic outbreak in the chi- 120904 (2019)
nese province hubei (2020). arXiv:2002.04482 25. Kuznetsov, Y., Piccardi, C.: Bifurcation analysis of periodic
11. van den Driessche, P., Watmough, J.: A simple sis epidemic seir and sir epidemic models. J. Math. Biol. 32, 109–121
model with a backward bifurcation. J. Math. Biol. 40, (1994)
525–540 (2000) 26. He, S., Banerjee, S.: Epidemic outbreaks and its control
12. Liu, J., Paré, P.E., Du, E., Sun, Z.: A networked sis disease using a fractional order model with seasonality and
dynamics model with a waterborne pathogen. 2019 Amer- stochastic infection. Physica A 501, 408-407 (2018)
ican Control Conference (ACC), pp. 2735–2740 (2019) 27. Wang, D., Tan, D., Liu, L.: Particle swarm optimization
13. Cai, Y., Kang, Y., Wang, W.: A stochastic sirs epidemic algorithm: an overview. Soft Comput. 22, 387–408 (2018)
model with nonlinear incidence rate. Appl. Math. Comput. 28. Peng, Y., Sun, K., He, S., Peng, D.: Parameter identification
305, 221–240 (2017) of fractional-order discrete chaotic systems. Entropy 21, 27
14. Al-Rahman El-Nor Osman, M., Adu, I.K., Yang, C.: A (2019)
simple seir mathematical model of malaria transmission. 29. Aguiar, M.S., Kooi, B.W., Stollenwerk, N.: Epidemiology
Asian Res. J. Math. 7, 1–22 (2017) of dengue fever: A model with temporary cross-immunity
15. Almeida, R.: Analysis of a fractional seir model with and possible secondary infection shows bifurcations and
treatment. Appl. Math. Lett. 84, 56–62 (2018) chaotic behaviour in wide parameter regions. Math. Model.
16. Chen, Y., Lu, P., Chang, C.: A time-dependent sir model for Natural Phenomena (2008)
COVID-19. (2020) 30. Sun, K., Liu, X., Zhu, C.: The 0–1 test algorithm for chaos
17. Tang, B., Wang, X., Li, Q., et al.: Estimation of the trans- and its applications. Chin. Phys. B 19(11), 110510 (2010)
mission risk of the 2019-ncov and its implication for public 31. He, S., Sun, K., Wang, H.: Complexity analysis and dsp
health interventions. J. Clin. Med. 9, 462 (2020) implementation of the fractional-order lorenz hyperchaotic
18. Wu, J.T., Leung, K., Leung, G.M.: Nowcasting and fore- system. Entropy 17, 8299–8311 (2015)
casting the potential domestic and international spread of 32. Costa, M., Goldberger, A.L., Peng, C.K.: Multiscale entropy
the 2019-ncov outbreak originating in wuhan, china: a analysis of complex physiologic time series. Phys. Rev.
modelling study. Lancet (Lond. Engl.) 395, 689–697 (2020) Lett. 6, 068102 (2002)
19. Wang, H., Wang, Z., Dong, Y., et al.: Phase-adjusted esti-
mation of the number of coronavirus disease 2019 cases in
Publisher’s Note Springer Nature remains neutral with
wuhan, china. Cell Discov. 6, 462 (2020)
regard to jurisdictional claims in published maps and
20. Salarieh, H.: Chaos analysis in attitude dynamics of a
institutional affiliations.
flexible satellite. Nonlinear Dyn. 93(3), 1421–1438 (2018)
21. Krysko, V.A.: Chaotic vibrations of flexible shallow axially
symmetric shells. Nonlinear Dyn. 91(4), 2271–2291 (2018)

123