Professional Documents
Culture Documents
A Model Based Study On The Dynamics of COVID-19 - Prediction and Control - Elsevier Enhanced Reader
A Model Based Study On The Dynamics of COVID-19 - Prediction and Control - Elsevier Enhanced Reader
0
4 Analysis of the system for fxed control
ln this section we assume the fixed value for control parameter
M, Here we mainly study the uniformly boundedness of the so-
lutions and subsequently the basic reproduction number, different
‘equilibria and their stability criteria, sensitivity ete.
3.1. Boundedness ofthe system
Here we examine the boundedness property of the system (1).
‘Theorem 3:1. The solutions of the system (1) are uniformly bounded,
Proof. We assume that X=S + FQ 4148
$
a
Therefore
ax
qt il=A—dx
ax
ie, Rd A
Integrating the above inequality and by applying the theorem of
differential equation due to Birkhoff and Rota [2], we get
[1] 4x0
Tow
Now for ¢ > c,
A
OS NS Tpit
Hence all the solutions of (1) that are initiating in {R§) are con-
fined in the region
Ate
Tp
for any ¢ > 0 and for t+ co, Hence the theorem, 0)
(Ke RE0=XE QR) <
32, Basie reproduction mumer
In any epidemic model, the basic reproduction number is the
‘most important epidemiological parameter for determining the na~
ture of a disease. Generally i is denoted by Ry and is defined as
the “number of secondary infected individuals caused by single in-
fected individuals in the whole time interval” (see, van den Driess-
‘che and Watmough [6]. Therefore the dimensionless quantity Ro
refers as the expectation of the spreading disease.
In literature, several techniques ate available to evaluate Ry for
‘an epidemic spread. In our present research article we use the next4 -M. Mondat, Jana and SK Noa cto Chas, Sotons and Fractal 1952020) 10889
generation matrix approach (5,922), Now the classes which are di
rectly involved for spread of disease is only &, Q I. Therefore from
system (1) we have
& — pa py past - nsf -et -0F de
dQ
42 _ 5.6 m9 coo @
a
Hak sansa soo.
The above system can be writen as $= O09) - WO).
p Athan)
wee y=(a). on (" "8 w=
i °
(taro+ae
(by +e+d)Q— Doe
(y 4d +S) 0k—eQ
is clear from the system wat Eo AG. atts
It is clear from the system (1) that Eo(zAqy.0.0.0. i)
is a disease free equilibrium. Now the Jacobian matrix of «>
and W at the disease free equilibrium are respectively given by,
(BC — pr pn)" 00}
(O16) = ° 0 0) and V=J(¥ Eo)
° 0 9
neatord 0 0
=p, btcrd 0
ma -¢ nsdsd
The basic reproduction number (Ro) isthe spectral radius of the
of the matrix (FV-') and for the present model itis given by
AB ~ pil = pa)
80 ip tava +a)
@)
33. Equilibria
The system has two possible equilibria, One is. disease
free equilibrium point. where infection vanishes from the sys-
tem. It is given by Eo(S", 0, 0, 0, 8), where $= zy
FO = zfiMg,. The other equilibrium point is ES", ,
1h) whese infection i abeays present in the system i
called endemic equilibria, where S* = 2M Fo = (by be
peptone ea ereeatirseaT
Dene i ea pain
p= lets cB peo egh ig ad)
Note I observed from the expression of the above tw equ
librium point is that te disease fee equlinum F aways ea
sible bt the endemic equilibrium Ey i Tease I Mp 1
34, Stability analysis
In the present section we investigate the local asymptotic sta-
bility criteria of the different equilibria,
Theorem 32. The disease free equilibrium Ey is locally asymptotically
stable if Ry < 1 and it is unstable if Ra > 1
Proof. The Jacobian matrix at the disease free equilibrium of the
system (1) is given by
asp) At pyyt pas? » oo
0” papier? -Grtare+e 0 oO
i] a tte) 00
it : ° ow
[Now the characteristic equation of the system (1) at its disease
free equilibrium is given by
C+D +d+ PMV +b) He+ d+ +d-45)
C4 (bp bar to +d) Ro) 4)
Clearly all the eigen value of the Jacobian matrix are negative
Jf and only if Ry = 1. Hence the system is locally asymptotically
stable if Ry = 1 and itis unstable if Rp > 1. Hence the theorem.
Note Here we see that the disease free equilibrium Ep losses its
stability when the Ry increases to its value greater than 1. So, we
‘may conclude that at Rp the system (1) passes through a bifurca-
‘ion around its disease free equilibrium which are discussed in the
next theorem,
‘Theorem 3.3. The system (1) passes trough a eranseritica bifurca-
tion around its disease free equilibrium whet! Ro = 1
Proof. From the above analysis, it has been observed that when
Rp = 1 between the two equilibria, only the disease free equi-
librium exists and locally asymptotically stable where as Ro > 1
fs the threshold condition for both existence and asymptotic sta-
bility criteria of the endemic equilibrium point although the dis
ease free equilibrium reduces to unstable nature at the threshold
Rp > |. Hence we may conclude that there is change of feasibility
as well as stability occurs at Rg = 1. Following the articles Guken-
hheimer and Holmes 12), Kar and Jana 21) Jana etal. (19) ee, itis
concluded that the system undergoes a transcrtical bifureation at
y= 1 and in Fig. 4, we represent graphically the phenomenon of
‘wanseviticl bifurcation,
Now we study the local asymptotic stability of the endemic
equilibrium £}
Theorem 3.4, The Endemic equilibrium F; is locally asymptotically
stable for Ro =
Proof. The jacobian matrix for the system (1) is given by
fy Bp pass oo
pet pte on 0 °
° by citer 00
° a © ~rdes) 0
ou ° o nd
where, an =—8C~ py) — p2)E*— (d+ PM), a = BO
py) = p2)S* = (bao +d), The characteristic equation of
the system (1) around its endemic equilibrium E, is
4A +m 444007 462 4GR46)=0 6)
where ¢, =2d-+b) +64 PM + BU px) pa)
C= (by + 6+ 0)(d-+ pM + BU — py) — pa)E") + (by tet
© + )BA — p)(1 ~ p29).
C= [bp Feta +d), +64 d) — DyDaI BA — py) ~ pF
It is clear from the equation (5) that first two roots are nega~
tive real numbers and remaining roots are the roots of the cubic
polynomial. It is also observe that here Cy, Cz, C3 and GC) ~C3 all
are postive for any parametric value. Hence following the Routh-
Hurwitz criterion we may conclude that the system (1) is locally
asymptotically stable around its endemic equilibrium Ey. 0
4, Optimal control problem
Here we focus on the time varying control M{t), which repre-
sents the awareness due to media coverage. Now it is very impor-
{ant to find out a strategy which minimizes the number of infected
persons as well as the associated cost. In this regard, the optimal
control theory is a very powerful tool to figure out such policy.
‘Therefore we consider the optimal control problem to minimize1M Mando Jona an SK None at/Chas Solitons and Frac 136 (202) 105880 5
the objective functional. Following [27.31], we construct the objec-
tive funetional as follows:
if "Teil + coho) 6)
‘subject to the proposed model (1), The parameters cy and ¢2 cor
responds as the weight constraints for the infected population and
the control respectively. Here the objective functional is linear in
the control with bounded states. Therefore it can be be showed by
using standard results that an optimal control and corresponding
‘optimal states exist (S|. Now we need to find out the value of the
‘optimal control BP) such that
JO") = ming)
Where @=(M(Q):0 0,
Now to investigate the singular case, we assume that
‘on some non-trivial interval. n this case we caleulate 2 (3)
After some simplifications of the time derivative of sf. we ob-
tain
a (aH) _a@
5(au) Feat Os baypS)
2 pS — Ay p+ FspS + eps. (12)
22) «assay By
+ (A—KSE + nQ — d5)(A5 — 21) + pdSis, (13)
‘occur in the above expression, so next we calculate the second
‘desivative with respect to time.
+ 8G)
= [W(E(i2 9a) + 02 — ae) a
(REQ — a) — dap
+ (A= R(SE +5) +10 — 08) sap
F(A KSE+ byQ — d5)(%s —7a)p-+ pdSHs + pdds (14)
where k= 8 — pyi(t~ po)
Using the state and co-state equations of systems (1) and (°)
‘we simplify the Eq. (14) and finally obtain
£()
ae (aw
= PPS(RE (as ~ 1) — dda) + (kPSE + dp8)(A5 ~ Aa).
+A ASE + byQ — d5)(25 ~)P ~ pes IMIC)
+ {ERS(A1 ~ Aa) + Ekigin +0 +d +a)
~ (byha + 04+ 09s )RE — BER — ha)
= UAE + RC — 2 RSE ~ (by 40-4 d+ EEN pS
~ (ADSKE + pS) + dpSKE Ag
+ (KE(0g ~ 31) ~ ddA ~KSE +10 — d5)P
+(A~RE(A—ISE + 5,Q - 5)
~ KSUSE — bE ~oE — dE ab)
+ bu(baE — (by +€-+d)Q) — ACA —KSE + 610 —d)]pC2s ~21)
+A ASE 4 byQ ~ dS) (dis ~ (KE 4 d)Ry + KERS)P + BSU?2s
+ pdas(A— KSE + by — dS) (15)
‘The above equation can be writen in the form
ps
KEp(As — br) +dp?S125— 1)
@ (aH
and then we can solve the singular control as
110 = [kpEOs
~ pPS(KECa2 — 91) ~ di) + (kpSE + dS) (is —
+ (A= KSE + byQ —AdS\(25 ~ 1) p— pSdas|
and
2(0) = (RESCH ~ fa) + Ekin(ba +0 +440)
= (aha + ha + 025 )KE — RE? (Aa — ha) — dk E
+h ~ Ba RSE (ba +d-+0 +0 E))pS
— (dpkSE + d?pS)Ay + dpkSEA2
+ (RE(Ag — 41) — dh (A — KSE + b,Q — dS)p
+1A~k(A~KSE + b,Q — dS)E
—KS(ISE— Eo — dE aE) + bytbab —(b; +€-44)Q)
~ (A= ASE +b, —d5)]pC2s 21)6 ‘M. Monat, Jaa and SK Nona cto hans, Sotons and Fatal 195 2020) 10889
F(AMKSE + b1Q —d5)(dis — (KE+d) 2
+ KEA,)P + DSPs + pag(A— KSE + b,Q — dS)
Moreover in order to satisfy the Generalized Legendre-clesh
Condition for the singular contol t© be optimal, we require
A Si(B) = 2100) to be negative [25]. Therefore we summarize
fhe Control profile on a nontrivial interval inthe following way:
a
<0, then MC)
em
an oS
it So, then Mee) =a,
aH oe)
iT Fy =O then Manin) = — Ft
Hence the control is optimal provided y(t) < 0 and a
~ tip =o.
5. Numerical simulation
We study numerical results in two different cases, frst for fixed
control and second when the control has been applied optimally.
First, we consider the values of parameters in Table 1, for numeri-
cal simulations. Since 5 is the disease induced mortality rate and
is the natural death rate, hence 8 > d, Using these parameters and
the initial conditions as S(O) = 500, £(0) = 10, Q(0) =5, (0)
1, R(0) =0, we solve numerically our proposed model. (1)
Fig, 2 verifies the numerical result when Ry < 1 and in
this case the solutions of the model (1) converge to the DFE
13 Q
Fig. 3. Dynamical behavior around FF,
‘he trisoia bration lag depicts the change of stably a
Ry ~ 1 and the corresponding solutions of the madel (1) converge
0 the EE, £,(S*,E*,Q°,1",R) = (48.16, 1.21, 15.73, 384, 16633),
‘These results are discussed in the Theorem 32 and Theorem 3.4
Further, in the Theorem 3.3, we have shown that the system (1)
‘may undergo through 2 transcritical bifurcation at the threshold
parametric condition Rp = 1. Therefore in Fig. 4, we demonstrate
the scenarios when Ry =1 and it has been observed that the
‘model (1) possesses a transeritical bifurcation there.
We now perform the sensitivity analysis to determine the
higher and lower impact of some parameters on the basic repro-
duction number. Ro and hence impact of such parameters on the
dynamics of the proposed model (1). To perform the sensitivity
analysis [22], we caleulate the normalized forward sensitivity in-
{ex of Ry and these indices measures the relative change in a Ro
with respect to the relative change in its parameters (Table 1).
Definition 5.1. The normalized forward sensitivity index of a func-
tion, Rx. #2. Ae} fOr (1 ones? os23769 090075 Estimated
_ ~ i Osr6i%9 —_Do2si61 Oot4s2 Estimated
bes
ig. 9. Vaation o the adjoin variables when canta apptied optima
J Ey]
ig. 10. Vaan ofthe contol strategy of cont parameter.
most populated country, has also been affected severely due to
the global pandemic COVID-19. In this paper, we analyze the ef-
fect of COVID-19 cases in three states in three different parts of
India. These states are the Western state Maharashtra (where the
‘commercial capital of India, Mumbai situated), the northern state
Delhi (where the capital of India, New Delhi is located), and the
southern state Tamil Nadu (where another highly populated mega-
city, Chennai is situated) and the some of rotal populations of these
three states crossing two hundred millions. We apply the model
system (1) to study the pandemic situation due to COVID-19 in
these three states of India from 2nd March, 2020 to 27th April,
2020 and predict the future behavior of the disease in a short term
basis. Although the first case of COVID-19 in India was confirmed
in the state Kerala on 30th January 2020, the frst cases of COVID-
19 were reported in the state of Delhi, Tamil Nadu, and Maharash-
tra respectively on 2nd, 7th, and 9th March 2020, The active, re-
covered, death and confirmed COVID-19 cases of these three states
are collected from the official websites of the oficial updates of
coronavirus, COVID-19 in India, Government of India 36], Govern-
meat of Delhi [11], Government of Maharashtca [10] and Govern-
Ina values of population inthe states of Mahara, Dethl and Tami
ad,
Tatil Vales Maharashra Deni Tail acs Reerences
50) TS TSSS 9ATOGO 10.0143)
£0) au 2a 3a Estimated
0) 98s as no Estimated
1 ° a 0. (roan.t436),
Ro ° a oa r436},
E 00
Fig Ave COVID-19 cases in Nahas,
‘ment of Tamil Nadu[ 4] and presented during this period in the
‘Table 3,
‘We fit the proposed model (1) to the daily active infected and
confirmed (cumulative) infected COVID-19 cases in those three
states of India using the set of parameters as given in Table 4 and
the intial size of the population from the Table 5. To fit these real
data, we use the software Mathematica and then predict the be-
havior of COVID-19 for those three states on a short term basis. In
Figs. 11-Fig. 12, we respectively present the active COVID-19 cases
in Maharashtra, Delhi, and Tamil Nadu for 91 days starting from
2nd March, 2020, till the 31st May 2020. Also, in Figs. 14, 15 and
in Fig. 16, we present the cumulative confirmed (ie. the sum of
active cases, recovered and death) COVID-19 cases of Maharashtra,
Delhi, and Tamil Nadu, respectively, for the same period
7. Discussion and conclusion
After world war Il, the globe has never undergone like the
present scenatios that arrived due to the pandemic COVID-19.
Starting from Wuhan city in December 2019, the virus SARS-CoV=
2 has spread throughout as many as 210 countries and territories
and continues to increase its pandemic nature. Inthe absence of
proper vaccination and treatment of COVID-19, one would rely on0 -M. Mondat, Jana and SK Noa cto Chas, Sotons and Fractal 1952020) 10889
Fi. 12, Active COMID-9 cases in Dei
Fig 18, Active COVID-I9 cases in Tat Na
qualitative control of the disease rather than complete eradication
Since these types of situations are unknown to present society, in
these circumstances, various governments have taken some poli-
cies ۩ control the destructive nature of the disease as much as
possible. To study the dynamics of the disease, in this article, we
have proposed and analyzed a classical SEIR type mathematical
‘model to incorporate the COVID-19 scenarios in the system (1)
[A detalled analysis shows that the proposed system posses two
equilibria, namely one disease-free and one endemic whose ex-
istence and asymptotic stability criteria depend on the numerical
Value of basic reproduction number Ro. We have also established
that the proposed system (1) undergoes a transcritical bifurcation
at the threshold Ry =. To study the impact of the government
Control measures (0 prevent the extensive transmission of COVID-
19, we have introduced a control parameter M, Further, to reduce
the infected individuals as well as to minimize the cost of im-
plementing government control measures. an objective functional
has been formed and solved using Pontryagin’s maximum princi-
ple. The optimal control follows a combination of bang-bang and
singular control during its application. We also have discussed the
sensitivity index ofthe threshold parameter fp and found the most
sensitive parameter. which has a positive impact on Rp is the dis-
ease transmission rate,
Next, we consider three cases of populations, namely (i) active
«cases of COVID-18, (ii) confirmed cases of COVID-18, and (ii) re-
covered cases of COVID-19 in three states of India, namely, western
sate Maharashtra, northern state Delhi, and southern state Tamil
Nadu where there are more than Gvo hundred people residing. In
‘Table 3, we provide the data of COVID-19 confirmed, active, recov-
ered, and death cases of these three states stating from 2nd March
’
3
:
Fig. 1. Contrmed COMD-19 eases in Manassa
Fig. 5. Coofemed COMD-9 eases in De
2020, till 27th April 2020, In Tables 4 and 5, we estimate the para-
‘metric Values associated with the model system (1). Finally, using
the software Mathematica, we try to fit our model (1) to estimate
the nature of COVID-19 in those three states of India. In Figs. 1
13, we present the active COVID-19 cases in Maharashtra, Delhi,
and Tamil Nadu, respectively, for 91 days starting from 2nd March
2020, wll the 31st of May 2020. In particular, we use the exist
ing data for the first 57 days (ull 27th April 2020) collecting from
the websites of the Government of India and WHO to forecast the
probable active COVID-19 cases in those three states for the rest
of 34 days, It should be noted that here we can forecast the na-
ture of the COVID-19 for the short term only as the Governmental
policy would change in time to time resulting in the correspond-
Ing changes in the associated parameters of the proposed model
system. From those three figures, we can claim that on 31st May
2020, the active cases of the three states Maharashtra, Delhi, and
‘Tamil Nadu, will be around 46183, 3344, and 411 respectively. Sim-
ilaly, in Figs. 14, 15 and in Fig. 16, we respectively present the cu-
‘mulative confirmed (ie. the sum of active, recovered and) death)
COVID-19 cases of Maharashtra, Delhi, and) Tamil Nadu starting
from 2nd March 2020, tll the 31st May 2020. Further, using the
existing data for the first 57 days (til 27th April 2020), the cases
for the next 34 days has been predicted. It has been forecasting
from those three figures that till 31st May 2020, the cumulative
confirmed instances of those three states Maharashtra, Delhi, and
‘Tamil Nadu will be around 83958, 19667, and 5133 respectively.
Moreover, from above figures (Fiz. 11 to 16), we can estimate the
‘number of active cases and number of cumulative confirmed eases
of those three states at any day of May, 2020,
The primary finding of this article is that we have derived
‘4 mathematical model that can be used to study the qualitative1M Mando Jona and SK Nan! ak/Chas Soto and Frac 1362020) 108880 0
eb 8
Fig. 16, Confirmed COMD-19 eases in Tail Nadu
‘dynamics of COVID-19. The basic reproduction number and its sen-
sitivity analysis would determine the controling procedure of the
disease. Also, we have incorporated the governmental policy in
‘our mathematical model and proposed a linear objective functional
considering that governmental policy as the time-dependent con-
trol parameter. On the other hand, it should be noted that the es-
timated values of COVID-19 in the three different states of India in
three different parts are forecasted using existing parametric space
‘only. In the case ofthe first two states, Maharashtra and Delhi we
see that the graph shows an increasing trend whereas for the state
Tamil Nadu, the predicted data shows that the disease can be con-
trolled with the existing parametric space, The main two burdens
behind the control of COVID-19 are (i) the unconsciousness of the
people about the disease and (ji) the high population density ex-
Posed the infection when the common people are out for their es-
‘ential commodities. Moreover, the weakness of forecasting cumu-
lative COVID-19 cases in the three states of India is that here the
Forecast is done on otally some existing parametric conditions. But
human behaviour is the most uncertain phenomena. Hence. if the
corresponding parametric space has been altered then there may
be some changes in the graphs of COVID-19 cases. Therefore, in
this article, we are targeting to forecast the COVID-19 cases on a
short term basis and henceforth there would be very litte chance
to change in corresponding parametric space. But the framework of
the current model gives some important insights into the dynam
ies of COVID-19 spread and control, Further, our simulation works
‘suggest that both the quarantine and governmental intervention
strategies like lockéown, media coverage on social distancing and
public hygiene can play an important role in diminishing COVID-
19 transmission, However, the successes of these strategies mainly
rely on the proper implementation of the process. In our future
work, we need to extend our model by incorporating some essen
tial biological as well as epidemiological factors. To be given some
‘examples, we may mention the acquired immunity in humans, dif-
ferential susceptibility and infectivity of humans to COVID-19 in-
fection, and spatial heterogeneity
Declaration of Competing Interest
‘The authors declare that they have no known competing finan-
ial interests or personal relationships that could have appeared to
Influence the work reported in this paper.
(CRediT authorship contribution statement
Manotosh Mandal: Conceptualization, Methodology, Software
Formal analysis, Investigation, Validation, Writing - original
‘raft. Soovoojeet Jana: Conceptualization, Validation, Methodol-
‘ogy, Writing - original draft, Writing - review & editing. Visual-
ization, Swapan Kumar Nandi: Validation, Formal analysis, Inves-
tigation. Anupam Khatua: Investigation, Resources, Data curation
Sayani Adak: Resources, Data curation, TK. Kar: Conceptualization,
‘Writing - review & editing, Visualization, Supervision,
‘Acknowledgment
“The work of, Jana and S. Adak are partially supported by Dept
of Science and Technology & Biotechnology, Govt. of West Ben-
gal (vide memo no, 201 (Sanc.)ST)PISRT/16G-12/2018 dated 19-
(02-2019). Research of A. Khatua is financially supported by De-
partment of Science and Technology-INSPIRE, Government of India
(No. DST/INSPIRE Felloship/2016/IF1G0667, dated: 21st Septem-
ber, 2016), Further the authors are very much grateful to the
‘anonymous reviewers and Prof. Gabriel Mindlin, Handling Editor
‘of the journal, for their constructive comments and useful sugges-
tions, which have helped us significant improvement of the article
References
10] BermoulD, Esa dune nowwelle analyse de ts morale cause pat Ls pte
ero: Ales Math iy Acad Ray Se Pas 1760, p 05.
(2) kat Cota GL. Ora difereal equations Boston Gin; 1082
1B] Buono B. Onto A LacignolaD. Cabal siabity of an SR epidemic
‘model mith formation dcpendent vaccination Math Bost 200816818.
(4) Charabrty Gosh Ream forests ane sk assessment of nove como
‘ivr (COMIG-0) cases a datardeven analy, Chive Stone Fractals 202
4101016 cnaos 202010850.
{51 Dekmann 0 JA), JAPH On the defetion on the computation o the nase
reproduction number ratio yin models for infectious dass i Reeroge-
‘as population. J Math Bol 1090:28:365-82
(6) Dresscne PV, Watmough J Rerecucton numbers and sub-thresolé endemic
quia for compartment mades of dice tansmisson. Math Bas
2hna:i0-29-48
10) Fano 0, Piazza F. Analysis and forecast of COMID-19 spreading In China, Katy
nd France Chaos Sotons Fractals 20213410076.
(a) Fleming Wat Rise! RW. Deterministic and stochatle optimal conto. New
Yorke Springer Veg: 175
[0] Fafa Gk, bers Nc IAP H. The wetapopuatin dynamics ofan infections
You might also like