1
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS)
ISSN (Print) 2313-4410, ISSN (Online) 2313-4402
Β© Global Society of Scientific Research and Researchers
http://asrjetsjournal.org/
Numerical Study of Kermack-Mckendrik SIR Model to
Predict the Outbreak of Ebola Virus Diseases Using Euler
and Fourth Order Runge-Kutta Methods
Md. Tareque Hossaina, Md. Musa Miahb*, Md. Babul Hossainc
aDepartment of Textile Engineering, City University, Birulia, Savar, Dhaka, Bangladesh
b,cDepartment of Mathematics, Mawlana Bhashani Science and Technology University, Santosh, Tangail-1902,
Bangladesh
aEmail: tareque.ms@gmail.com, bEmail: musa_ju69@yahoo.com, cEmail: babulhossainh@yahoo.com
Abstract
Mathematical Modeling has emerged as a vital tool for understanding the dynamics of the spread of many
infectious diseases, one amongst is Ebola virus. The main focus of this paper is to model mathematically the
transmission dynamics of Ebola virus. For this purpose we tend to use basic SIR model of Ebola Virus to predict
the outbreak of the diseases. As we cannot fully solve the 3 basic equations of SIR model with a certain formula
solution, we introduce Euler and fourth-order Runge-Kutta methods (RK4). These two proposed strategies are
quite efficient and practically well suited for solving initial value problem (IVP) for ordinary differential
equations (ODE).We discuss the numerical comparisons between Euler method and Runge-Kutta methods and
also discuss regarding their performances with the actual data. The population that we used for this model had
roughly a similar number of individuals as the number was living in Republic of Liberia during 2014.
Keywords: Ebola; Outbreak; Evolution; Mathematical Modeling.
1. Introduction
Mathematical models are a powerful tool for investigating human infectious diseases, such as Ebola virus,
contributing to the understanding of the dynamics of disease and providing useful predictions about the potential
transmission of a disease and the effectiveness of possible control measures, which can provide valuable
information for public health policy makers.
------------------------------------------------------------------------
* Corresponding author.
http://asrjetsjournal.org/
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
2
The Ebola virus disease was first discovered in 1976 in the present Democratic Republic of Congo [1]. Since
then, there have been many outbreaks; with the greatest was 2014 outbreak [2] which has spread through many
countries. According to Bekoe (2015), from the first confirmed case recorded on 23 March 2014 which more
than 18 months, at least 11,312 people have been reported died from the disease in six countries; Liberia, Sierra
Leone, Guinea, Nigeria, Mali and US. Since the virus keeps spreading through contacts and the mortality rate of
0.7[3], it is needed to understand the patterns and epidemiology of the disease. By these conditions,
mathematical model of the outbreaks of Ebola virus can be helpful as it is a platform for understanding the
behavior of a dynamical system. The objectives are to understand better the mathematical dynamics of an
infected population when an outbreak occurs. Another goal is to use mathematical modeling to examine and to
analyze the viral dynamics of the Ebola virus. To model this outbreak, the systems of differential equations are
used. To study the known data, several distinct models will be used and each model is different depends on the
parameters acquired. I decided to model the Ebola Epidemics in Liberia in 2014 and compare their spread using
an SIR model.
1.1 Objectives
The purposes of this research are
1. To apply SIR model to predict the outbreaks of Ebola virus.
2. To determine the effect of the initial number of infectives of the population.
3. Compare with real data and fit with model.
1.2 Scope
In the proposal study, we will only investigate the outbreaks of Ebola virus by applying the mathematical model.
The data has been collected and it covers the area in the continent of Africa, on Liberia that was recorded by
CDC [4]. The calculation will be done by using tools of MATLAB software.
2. Methodology
2.1 Formulation of SIR model
The SIR model is used to illustrate the transfer of the epidemic through the interaction of the following three
different variables:
ππ = Number of people
that are susceptible to Ebola
πΌπΌ = Number of people infected with Ebola
π
π
= Number of people recovered
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
3
from Ebola with total immunity
It makes sense to assume that a fixed population of ππ people, whereby there are no births and deaths by natural
cause i.e.
ππ = ππ + πΌπΌ + π
π
[5]
This is because the population is fixed and therefore, there are only three compartments in which the population
may fit into. Thus, the total of the number of people susceptible infected and recovered in equivalent to the total
population. The assumption that ππ is fixed, with no births or deaths, makes sense given 60 days, although it is a
simplification.
These variables change over time, so we will define the variable π‘π‘ = time in days. We will set π‘π‘ = 0 at the start
of August 2014.
The model uses two parameters π½π½ and πΎπΎ with π½π½, πΎπΎ > 0.
Given these parameters, the model uses 3 differential equations.
The rate of change of the number of people susceptible to the disease over time
ππππ
ππππ
= β π½π½πΌπΌππ (1)
The rate of change of the number of people recovered over time
ππππ
ππππ
= πΎπΎπΌπΌ (2)
The rate of change of the number of people infected.
ππππ
ππππ
= π½π½πΌπΌππ β πΎπΎπΌπΌ (3)
Parameterization of the model
In order to calculate π½π½ (the rate of infection) and πΎπΎ (the rate of recovery), it helps to define two more parameters.
π·π· = Duration of disease for those recovered
ππ = Mortality rate for those who die per day
(0.7 for Ebola)
This leads to two further equations.
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
4
The rate at which the disease is spread
πΎπΎ = 1
π·π·
[6] (4)
The infection rate of the disease
π½π½ = ππ
ππ
[7] (5)
2.2 Transformation of Runge-Kutta Equations for SIR modeling
RK4 is one of the classic methods for numerical integration of ODE models.
Consider the following initial problem of ODE
ππππ
ππππ
= ππ(π‘π‘,π¦π¦)
π¦π¦(π‘π‘ππ) = π¦π¦ππ
Where π¦π¦(π‘π‘) is the unknown function (scalar or vector) which I would like to approximate.
The Iterative formula of RK4 method for solving ODE is as follows
ππ1 = βππ(π‘π‘ππ,π¦π¦ππ)
ππ2 = βππ οΏ½π‘π‘ππ +
1
2
β,π¦π¦ππ +
1
2
ππ1οΏ½
ππ3 = βππ οΏ½π‘π‘ππ + 1
2
β,π¦π¦ππ + 1
2
ππ2οΏ½
ππ4 = βππ(π‘π‘ππ + β,π¦π¦ππ + ππ3)
π¦π¦ππ+1 = π¦π¦ππ +
1
6
(ππ1 + 2ππ2 + 2ππ3 + ππ4)
For simplicity, here we use the simplest SIR model to examine whether the RK4 method has been implemented
correctly. The SIR model is defined as follows
ππππ
ππππ
= β π½π½πΌπΌππ
ππππ
ππππ
= π½π½πΌπΌππ β πΎπΎπΌπΌ
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
5
ππππ
ππππ
= πΎπΎπΌπΌ
where ππ(π‘π‘) is the number of susceptible people in the population at time π‘π‘, πΌπΌ(π‘π‘) is the number of infectious
people at time π‘π‘, π
π
(π‘π‘) is the number of recovered people at time π‘π‘, π½π½ is the transmission rate, πΎπΎ represents the
recovery rate, and
ππ = ππ(π‘π‘) + πΌπΌ(π‘π‘) + π
π
(π‘π‘) is the fixed population.
According to the general iterative formula, the iterative formulas for ππ(π‘π‘), πΌπΌ(π‘π‘) and π
π
(π‘π‘) of SIR model can be
written out
ππππ+1 = ππππ +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
ππ1ππ = ππ(π‘π‘ππ, ππππ , πΌπΌππ) = βπ½π½πππππΌπΌππ
ππ2ππ = ππ οΏ½π‘π‘ππ + βππ
2
, ππππ + ππ1
ππβππ
2
, πΌπΌππ + ππ1
πΌπΌβππ
2
οΏ½
= βπ½π½ οΏ½ππππ +
ππ1ππβπ‘π‘
2
οΏ½ (πΌπΌππ +
ππ1ππβπ‘π‘
2
)
ππ3ππ = ππ οΏ½π‘π‘ππ +
βπ‘π‘
2
, ππππ +
ππ2ππβπ‘π‘
2
, πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½
= βπ½π½ οΏ½ππππ +
ππ2ππβπ‘π‘
2
οΏ½ (πΌπΌππ +
ππ2ππβπ‘π‘
2
)
ππ4ππ = ππ(π‘π‘ππ + βπ‘π‘, ππππ + ππ3ππβπ‘π‘, πΌπΌππ + ππ3ππβπ‘π‘)
= βπ½π½(ππππ + ππ3ππβπ‘π‘)(πΌπΌππ + ππ3ππβπ‘π‘)
πΌπΌππ+1 = πΌπΌππ +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
ππ1ππ = ππ(π‘π‘ππ, ππππ, πΌπΌππ) = π½π½πππππΌπΌππ β πΎπΎπΌπΌππ
ππ2ππ = ππ οΏ½π‘π‘ππ +
βπ‘π‘
2
, ππππ +
ππ1ππβπ‘π‘
2
, πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½ = π½π½ οΏ½ππππ +
ππ1ππβπ‘π‘
2
οΏ½οΏ½πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½ β οΏ½πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½
ππ3ππ = ππ οΏ½π‘π‘ππ +
βπ‘π‘
2
, ππππ +
ππ2ππβπ‘π‘
2
, πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
6
= π½π½ οΏ½ππππ +
ππ2ππβπ‘π‘
2
οΏ½οΏ½πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½ β οΏ½πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½
ππ4ππ = ππ(π‘π‘ππ + βπ‘π‘, ππππ + ππ3ππβπ‘π‘, πΌπΌππ + ππ3ππβπ‘π‘)
= π½π½(ππππ + ππ3ππβπ‘π‘)(πΌπΌππ + ππ3ππβπ‘π‘) β (πΌπΌππ + ππ3ππβπ‘π‘)
π
π
ππ+1 = π
π
ππ +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
ππ1ππ = ππ(π‘π‘ππ, πΌπΌππ) = πΎπΎπΌπΌππ
ππ2ππ = ππ οΏ½π‘π‘ππ +
βπ‘π‘
2
, πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½ = πΎπΎ οΏ½πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½
ππ3ππ = ππ οΏ½π‘π‘ππ +
βπ‘π‘
2
, πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½ = πΎπΎ οΏ½πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½
ππ4ππ = ππ(π‘π‘ππ + βπ‘π‘, πΌπΌππ + ππ3ππβπ‘π‘) = πΎπΎ(πΌπΌππ + ππ3ππβπ‘π‘)
2.3 Transformation of Euler Equations for SIR modeling
If we have a "slope formula," i.e., a way to calculate πππ¦π¦/πππ‘π‘ at any point (π‘π‘,π¦π¦), then we can generate a
sequence of π¦π¦-values,
π¦π¦0 ,π¦π¦1,π¦π¦2,π¦π¦3 β¦ β¦ β¦ β¦
By starting from a given π¦π¦0 and computing each rise as slope x run. That is,
π¦π¦ππ+1 = π¦π¦ππ + π π π π π π π π π π ππ βπ‘π‘
where βπ‘π‘ is a suitably small step size in the time domain.
It really doesn't matter in this calculation if the slope formula happens to depend not just on t and y but on other
variables, say x and z -- as long as we know how x and z are related to t and y. If x and z happen to be other
dependent variables in a system of differential equations, I can generate values of x and z in the same way.
Of course, for the SIR model, we want the dependent variable names to be ππ, πΌπΌ ππππππ π
π
. Thus we have three
Euler formulas of the form
ππππ+1 = ππππ + π π π π π π π π π π ππ βπ‘π‘
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
7
πΌπΌππ+1 = πΌπΌππ + π π π π π π π π π π ππ βπ‘π‘
π
π
ππ+1 = π
π
ππ + π π π π π π π π π π ππ βπ‘π‘
More specifically, given the SIR equations,
ππππ
ππππ
= β π½π½πΌπΌππ ; ππππ
ππππ
= π½π½πΌπΌππ β πΎπΎπΌπΌ; ππππ
ππππ
= πΎπΎπΌπΌ
The Euler formulas become
ππππ+1 = ππππ β π½π½πΌπΌππππππβπ‘π‘
πΌπΌππ+1 = πΌπΌππ + (π½π½πΌπΌππππππ β πΎπΎπΌπΌππ)βπ‘π‘
π
π
ππ+1 = π
π
ππ + πΎπΎπΌπΌππβπ‘π‘
To calculate something from these formulas, we must have explicit values for
π½π½, πΎπΎ, ππ(0), πΌπΌ(0), π
π
(0) ππππππ βπ‘π‘ .
3. Application, Comparison and Result Discussion
If we now take the example of the Ebola outbreak in Liberia 2014, we can assign the parameters with the
following values. The total population of Liberia, N = 4294000 [8], and according to data from WHO, the
number of people infected, I =391[9] and the number of people dead is 227 [9]. Seeing as R includes the number
of people who have received permanent immunity, this includes those who have died as they have permanent
immunity, in addition to those who have recovered with permanent immunity.
Therefore, number of people recovered π
π
= 227 + (0.3 Γ 391) = 344
We will now use this data to provide the parameters with the following values.
ππ = 4294000; πΌπΌ = 391; π
π
= 344
Therefore, ππ = ππ β πΌπΌ + π
π
= 4294000 β (344 + 391) = 4293265
The duration of the disease ranges from 2 to 21 days [10], therefore we could roughly estimate the duration of
the disease at the midpoint, i.e.11 (approx.) days.
π·π· = 11; πΎπΎ = 1
11
= 0.09
According to WHO, the mortality rate of Ebola is 0.7 [3] and the number of people susceptible is 4293265.
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
8
Therefore, π½π½ (the rate of infection) = 0.7
4293265
= 1.63 Γ 10β7
In order to use the SIR model to predict the evolution of the disease, it would be helpful if we could solve the
system of differential equations.
Unfortunately, we cannot completely solve these equations with an explicit formula solution.
Therefore, we will use numerical approaches. We will use Euler method and fourth order Runge-Kutta method
to extract the solution.
3.1 Euler Method
For each day, we will calculate the values of ππ, πΌπΌ and π
π
using
1. ππππ+1 = ππππ β π½π½πΌπΌππππππ
2. πΌπΌππ+1 = πΌπΌππ + (π½π½πΌπΌππππππ β πΎπΎπΌπΌππ)
3. π
π
ππ+1 = π
π
ππ + πΎπΎπΌπΌππ
We take the initial values as
ππ0 = 4293265; πΌπΌ0 = 391; π
π
0 = 344; πΎπΎ = 0.09 ; π½π½ = 1.63 Γ 10β7;
We will do this explicitly for the transition from t = 0 to t = 1. Using equations 1, 2 and 3, the following values
for S, I and R can be calculated.
ππ1 = ππ0 β π½π½πΌπΌ0ππ0 Γ βπ‘π‘ = 4293265 β ( 1.63 Γ 10β7 Γ 391 Γ 4293265) Γ 1
= 4292991.377 β 4292991
πΌπΌ1 = πΌπΌ0 + (π½π½πΌπΌ0ππ0 β πΎπΎπΌπΌ0) Γ βπ‘π‘ = 391 + (1.63 Γ 10β7 Γ 391 Γ 4293265 β 0.09 Γ 391) Γ 1
= 629.4326582 β 630
π
π
1 = π
π
0 + πΎπΎπΌπΌ0
= 344 + 0.09 Γ 391
= 379.19 β 379
Here we use MATLAB to evaluate S, I, R over a two month period. Table-3.1 shows the result.
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
9
Table 3.1: Euler Method
Day S I
R
S+I+R
01 4293265 391 344 4294000
02 4292991 630 379 4294000
03 4292550 1014 436 4294000
04 4291842 1630 528 4294000
05 4290700 2625 675 4294000
06 4288865 4225 910 4294000
07 4285912 6798 1290 4294000
08 4281162 10936 1902 4294000
09 4273530 17584 2886 4294000
10 4261283 28248 4469 4294000
11 4241662 45327 7011 4294000
12 4210323 72586 11091 4294000
13 4160509 115868 17623 4294000
14 4081931 184017 28052 4294000
15 3959494 289892 44614 4294000
16 3772399 450898 70703 4294000
17 3495141 687575 111284 4294000
18 3103424 1017410 173166 4294000
19 2588759 1440508 264733 4294000
20 1980912 1918710 394378 4294000
21 1361382 2365556 567062 4294000
22 836453 2677585 779962 4294000
23 471386 2801669 1020945 4294000
24 256117 2764788 1273095 4294000
25 140695 2631379 1521926 4294000
26 80349 2454901 1758750 4294000
27 48198 2266111 1979691 4294000
28 30394 2079964 2183642 4294000
29 20090 1903072 2370838 4294000
30 13858 1738028 2542114 4294000
31 9932 1585531 2698537 4294000
32 7365 1445400 2841235 4294000
33 5630 1317049 2971321 4294000
34 4422 1199723 3089855 4294000
35 3557 1092613 3197830 4294000
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
10
36 2923 994911 3296166 4294000
37 2449 905843 3385708 4294000
38 2088 824679 3467233 4294000
39 1807 750738 3541455 4294000
40 1586 683393 3609021 4294000
41 1409 622065 3670526 4294000
42 1266 566222 3726512 4294000
43 1150 515378 3777472 4294000
44 1053 469091 3823856 4294000
45 972 426953 3866075 4294000
46 905 388595 3904500 4294000
47 847 353679 3939474 4294000
48 799 321896 3971305 4294000
49 757 292968 4000275 4294000
50 720 266637 4026643 4294000
51 689 242671 4050640 4294000
52 662 220858 4072480 4294000
53 638 201004 4092358 4294000
54 617 182935 4110448 4294000
55 599 166489 4126912 4294000
56 583 151521 4141896 4294000
57 568 137899 4155533 4294000
58 555 125501 4167944 4294000
59 544 114217 4179239 4294000
60 533 103948 4189519 4294000
3.2 Runge-Kutta (RK4) Method
For each day, we will calculate the values of ππ, πΌπΌ and π
π
using
ππ1ππ = βπ½π½πππππΌπΌππ
ππ1ππ = π½π½πππππΌπΌππ β πΎπΎπΌπΌππ
ππ1ππ = πΎπΎπΌπΌππ
ππ2ππ = βπ½π½ οΏ½ππππ +
ππ1ππβπ‘π‘
2
οΏ½ (πΌπΌππ +
ππ1ππβπ‘π‘
2
)
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
11
ππ2ππ = π½π½ οΏ½ππππ +
ππ1ππβπ‘π‘
2
οΏ½ οΏ½πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½ β οΏ½πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½
ππ2ππ = πΎπΎ οΏ½πΌπΌππ +
ππ1ππβπ‘π‘
2
οΏ½
ππ3ππ = βπ½π½ οΏ½ππππ +
ππ2ππβπ‘π‘
2
οΏ½ (πΌπΌππ +
ππ2ππβπ‘π‘
2
)
ππ3ππ = π½π½ οΏ½ππππ +
ππ2ππβπ‘π‘
2
οΏ½ οΏ½πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½ β οΏ½πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½
ππ3ππ = πΎπΎ οΏ½πΌπΌππ +
ππ2ππβπ‘π‘
2
οΏ½
ππ4ππ = βπ½π½(ππππ + ππ3ππβπ‘π‘)(πΌπΌππ + ππ3ππβπ‘π‘)
ππ4ππ = π½π½(ππππ + ππ3ππβπ‘π‘)(πΌπΌππ + ππ3ππβπ‘π‘) β (πΌπΌππ + ππ3ππβπ‘π‘)
ππ4ππ = πΎπΎ(πΌπΌππ + ππ3ππβπ‘π‘)
ππππ+1 = ππππ +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
πΌπΌππ+1 = πΌπΌππ +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
π
π
ππ+1 = π
π
ππ +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
We take the initial values as
ππ0 = 4293265 ; πΌπΌ0 = 391; π
π
0 = 344 ; πΎπΎ = 0.09; π½π½ = 1.63 Γ 10β7
We will do this explicitly for the transition from t = 0 to t = 1. Using those equations the following values for S,
I and R can be calculated.
ππ1ππ = βπ½π½ππ0πΌπΌ0
= β1.63 Γ 10β7 Γ 4293265 Γ 391 = β273.622658245000
ππ1ππ = π½π½ππ0πΌπΌ0 β πΎπΎπΌπΌ0 = 1.63 Γ 10β7 Γ 4293265 Γ 391 β 0.09 Γ 391
= 238.432658245000
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
12
ππ1ππ = πΎπΎπΌπΌ0 = 0.09 Γ 391 = 35.1900000000000
ππ2ππ = βπ½π½ οΏ½ππ0 +
ππ1ππβπ‘π‘
2
οΏ½ (πΌπΌ0 +
ππ1ππβπ‘π‘
2
)
= β1.63 Γ 10β7 Γ οΏ½4293265 β
273.622658245000
2
οΏ½ Γ οΏ½391 +
238.432658245000
2
οΏ½
= β357.039129094785
ππ2ππ = π½π½ οΏ½ππ0 +
ππ1ππβπ‘π‘
2
οΏ½οΏ½πΌπΌ0 +
ππ1ππβπ‘π‘
2
οΏ½ β οΏ½πΌπΌ0 +
ππ1ππβπ‘π‘
2
οΏ½
= β1.63 Γ 10β7 Γ οΏ½4293265 +
β273.622658245000
2
οΏ½ Γ οΏ½391 +
238.432658245000
2
οΏ½
β οΏ½391 +
238.432658245000
2
οΏ½ = 311.119659473760
ππ2ππ = πΎπΎ οΏ½πΌπΌ0 + ππ1
πΌπΌβππ
2
οΏ½ = 0.09 Γ οΏ½391 + 238.432658245000
2
οΏ½ =45.9194696210250
ππ3ππ = βπ½π½ οΏ½ππ0 +
ππ2ππβπ‘π‘
2
οΏ½ (πΌπΌ0 +
ππ2ππβπ‘π‘
2
)
= β1.63 Γ 10β7 Γ οΏ½4293265 +
β357.039129094785
2
οΏ½ οΏ½391 +
311.119659473760
2
οΏ½
= β382.467864374178
ππ3ππ = π½π½ οΏ½ππ0 + ππ2
ππβππ
2
οΏ½ οΏ½πΌπΌ0 + ππ2
πΌπΌβππ
2
οΏ½ β οΏ½πΌπΌ0 + ππ2
πΌπΌβππ
2
οΏ½
= 1.63 Γ 10β7 Γ οΏ½4293265 +
β357.039129094785
2
οΏ½ Γ οΏ½391 +
311.119659473760
2
οΏ½
β οΏ½391 +
311.119659473760
2
οΏ½
= 333.277479697859
ππ3ππ = πΎπΎ οΏ½πΌπΌ0 + ππ2
πΌπΌβππ
2
οΏ½ = 0.09 Γ οΏ½391 + 311.119659473760
2
οΏ½ = 49.1903846763192
ππ4ππ = βπ½π½(ππ0 + ππ3ππβπ‘π‘)(πΌπΌ0 + ππ3ππβπ‘π‘)
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
13
= β1.63 Γ 10β7 Γ (4293265 β 382.467864374178) Γ (391 + 333.277479697859)
= β506.805816985307
ππ4ππ = π½π½(ππ0 + ππ3ππβπ‘π‘)(πΌπΌ0 + ππ3ππβπ‘π‘) β (πΌπΌ0 + ππ3ππβπ‘π‘)
= 1.63 Γ 10β7 Γ (4293265 β 382.467864374178) Γ (391 + 333.277479697859)
β (391 + 333.277479697859)
= 441.62084381250
ππ4ππ = πΎπΎ(πΌπΌ0 + ππ3ππβπ‘π‘)
= 0.09 Γ (391 + 333.277479697859) = 65.1849731728073
β΄ ππ1 = ππ0 +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
= 4293265 +
1
6
(β273.622658245000 β 2 Γ 357.039129094785 β 2 Γ 382.467864374178
β 506.805816985307)
= 4292888.42625631 β 4292888
β΄ πΌπΌ1 = πΌπΌ0 +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
= 391 +
1
6
(238.432658245000 + 311.119659473760 + 333.277479697859 + 441.62084381250)
= 719.141296733456 β 719
β΄ π
π
1 = π
π
0 +
βπ‘π‘
6
(ππ1ππ+2ππ2ππ+2ππ3ππ+ππ4ππ)
= 344 +
1
6
(35.1900000000000 + 45.9194696210250 + 49.1903846763192 + 65.1849731728073)
= 392.432446961249 β 393
Here we use MATLAB to evaluate S, I, R over a two month period.Table-3.2 shows the result.
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
14
Table 3.2
RK4 Method
Day S I R S+I+R
01 4293265 391 344 4294000
02 4292888 719 393 4294000
03 4292196 1322 482 4294000
04 4290923 2432 645 4294000
05 4288583 4471 946 4294000
06 4284286 8214 1500 4294000
07 4276407 15077 2516 4294000
08 4261994 27626 4380 4294000
09 4235753 50457 7790 4294000
10 4188376 91623 14001 4294000
11 4104130 164649 25221 4294000
12 3958283 290515 45202 4294000
13 3717118 496967 79915 4294000
14 3346977 809182 137841 4294000
15 2838838 1226389 228773 4294000
16 2237427 1696343 360230 4294000
17 1636735 2124419 532846 4294000
18 1127150 2428029 738821 4294000
19 747957 2580728 965315 4294000
20 489487 2605010 1199503 4294000
21 321556 2540850 1431594 4294000
22 214459 2424226 1655315 4294000
23 146134 2280695 1867171 4294000
24 102037 2126431 2065532 4294000
25 73073 1971024 2249903 4294000
26 53657 1819893 2420450 4294000
27 40360 1675938 2577702 4294000
28 31056 1540566 2722378 4294000
29 24413 1414309 2855278 4294000
30 19575 1297197 2977228 4294000
31 15987 1188973 3089040 4294000
32 13279 1089223 3191496 4294000
33 11204 997456 3285340 4294000
34 9589 913148 3371263 4294000
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
15
35 8316 835770 3449914 4294000
36 7300 764806 3521894 4294000
37 6479 699764 3587757 4294000
38 5809 640176 3648015 4294000
39 5257 585604 3703139 4294000
40 4798 535640 3753562 4294000
41 4414 489905 3799681 4294000
42 4089 448050 3841861 4294000
43 3813 409750 3880437 4294000
44 3577 374709 3915714 4294000
45 3374 342652 3947974 4294000
46 3199 313328 3977473 4294000
47 3046 286506 4004448 4294000
48 2913 261974 4029113 4294000
49 2796 239538 4051666 4294000
50 2694 219019 4072287 4294000
51 2603 200255 4091142 4294000
52 2523 183095 4108382 4294000
53 2452 167404 4124144 4294000
54 2389 153057 4138554 4294000
55 2333 139937 4151730 4294000
56 2283 127941 4163776 4294000
57 2238 116972 4174790 4294000
58 2197 106943 4184860 4294000
59 2161 97773 4194066 4294000
60 2128 89389 4202483 4294000
Figure 3.1-3.3 shows the comparison between Euler and RK-4 method in the cases of infectives, recovered and
susceptives respectively.
Figure 3.1
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
16
Figure 3.2
Figure 3.3
For predict the evolution, we only discuss about the infectives.
Each strategies show that initially, the number of individuals infected will increase steeply, however, over an
extended period of time, the numbers eventually decrease. This happens simultaneously because the range of
individuals recovered will increase because as those infected decreases, they're being transferred into the
recovered class. The main reason is owing to the raised awareness of the disease leasing to additional medical
support being given in order to assist combat the transmission of the disease. Furthermore, an increased
awareness ends up in additional folks being conscious of strategies of protection. The steep increase within the
beginning of the primary fifteen days is possibly to flow from to the good uncertainty that lied with Ebola
allowing a greater rate of transmission. The peak of every graph illustrates the utmost number of individuals
ever to be infected and once this point; there's a transition whereby the numbers decrease.
3.3 Comparing the model to actual data
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
17
However, in order for the model to be valid and allow informing government policy, it obviously needs to
correspond fairly close to reality.
The Table-3.3.1 below compares the data collected from the SIR model (using Euler and RK4) for the number
of people infected and the real life data of the number of people infected.
Table 3.3.1
Time(Days) I -Actual I- Euler I -RK4 Time(Days) I -Actual I- Euler I -RK4
1 391 391 391 35 1871 1092613 835770
2 486 630 719 37 2046 905843 699764
6 554 4225 8214 41 2081 622065 489905
10 599 28248 91623 45 2407 426953 342652
11 670 45327 164649 47 2710 353679 286506
13 786 115868 496967 51 3022 242671 200255
17 834 687575 2124419 53 3280 201004 167404
19 972 1440508 2580728 55 3458 166489 139937
20 1082 1918710 2605010 60 3696 103948 89389
26 1378 2454901 1819893
Using the value of Table-3.3.1 we can plot a graph using MATLAB which compares the model data to the
actual data for the number of people infected (Figure-3.4).
Figure 3.4
As the graph demonstrates, the real data does not correspond very well to the data received from the model.
Although the actual data may seem to follow a straight like graph, this is untrue as it is only depicted in this
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
18
manner due to the limitations on the axis of the graph. The difference between the real life data and the data
from the model is so vast that the straight line looks like a graph of π¦π¦ = 0. Therefore, we decided to plot it
separately (Figure-3.5).
Figure 3.5
3.4 Curve Fitting with Actual Data
The model has significantly overestimated the number of individuals who can become infected with Ebola
fever. This is often because of the many limitations that the model presents. One of the main limitations includes
the incorrect beta and gamma values that were calculated. Once fixing the beta and gamma values, we were able
to find another gamma value that resulted in similar values to the real data. Here is that the graph to indicate
this, with the suitable gamma value of 0.66. The Table-3.4.1 below compares the data collected from the fitted
model for the number of people infected and the real life data of the number of people infected.
Table-3.4.1
Time(Days) I -Actual I- Euler I -RK4 Time(Days) I -Actual I- Euler I -RK4
1 391 391 391 35 1871 1555 1419
2 486 408 407 37 2046 1679 1525
6 554 482 475 41 2081 1952 1758
10 599 569 555 45 2407 2261 2020
11 670 593 576 47 2710 2430 2163
13 786 644 623 51 3022 2796 2472
17 834 759 726 53 3280 2994 2639
19 972 824 783 55 3458 3201 2814
20 1082 858 813 60 3696 3757 3287
26 1378 1092 1019
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
19
Using the value of Table-3.4.1 we can plot a graph using MATLAB which compares curve fitted data with real
data for the number of people infected (Figure-3.6).
Figure 3.6
3.5 Result Discussion
Although, this doesn't match the graph exactly, it shows a much better correlation of the number of individuals
infected. Between Euler and RK4 technique, Rk4 provides better correlation as because Rk4 methods divide its
step size more than Euler methods and the local truncation error is less than Euler methods. Therefore, in order
to improve the model, many changes should be done, together with altering the gamma value. The value that we
eventually used to alter the model, led to being in 2 decimal places. This goes to illustrate the necessary
precision required as very little deviance will cause massive changes. This can be because; the gamma is
calculated through extreme simplification, leaving great possibilities for more room for errors. Furthermore, it is
tough to differentiate between the numbers of people who have died and also the numbers of people who have
survived with permanent immunity as they both fall under the same class of being βrecoveredβ. It is also a fact
that rate of recovery is faster that the time scale of birth and death. The infection rate and recovery rate plays a
vital role in the model and in this method we find more accurate value of infection rate and recovery rate to fit
the model with the actual data collected from CDC.
4. Conclusion
The results obtained from modeling data will lead to completely different views and interpretations. This is due
to the unequal distribution of knowledge across the globe whereby in countries like Liberia, there is little access
to the statistics which makes it troublesome to form constructive predictions regarding the outbreak. Through
our research, we have gained further insight into the uses of mathematical modeling so as to work out the
outbreak of diseases similarly as evaluating its flaws. Having chosen Ebola as the diseases of concentration, as it
is incredibly relevant to this situation in continent, it has enabled a practical understanding of its rate of
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
20
transmission.
5. Recommendations
The Numerical methods give a better prediction if the data collected from an authentic source. This result can
help to estimate future predictions of the disease and consequently, it will facilitate to determine practical
components like the quantity of beds required in the hospitable, number of vaccination and reallocation costs
etc.
References
[1] "Ebola Virus Disease". World Health Organization. N.p., 2017. Web. 29 Mar. 2017.
[2] "2014-2016 Ebola Outbreak In West Africa| Ebola Hemorrhagic Fever | CDC". Cdc.gov. N.p., 2017.
Web. 29 Mar. 2017.
[3] "WHO Finds 70 Percent Ebola Mortality Rate". Aljazeera.com. N.p., 2017. Web. 29 Mar. 2017.
[4] "Previous Case Counts| Ebola Hemorrhagic Fever | CDC". Cdc.gov. N.p., 2017. Web. 29 Mar. 2017.
[5] Dolgoarshinnykh, R. G., & Lalley, S. P. (2003). Epidemic Modelling: SIRS Models (Doctoral
dissertation, University of Chicago, Department of Statistics).
[6] "Modelling Infectious Diseases." IB Maths Resources From British International School Phuket".
ibmathsresources.com. N.p., 2017. Web. 29 Mar. 2017.
[7] "The Spread Of Infectious Diseases." The British Medical Journal 2.1281 (1885): 108. Web.
[8] Leone, S. Appendix: Additional Results and Technical Notes for the EbolaResponse Modeling Tool
Additional Results. Population, 4, 3.
[9] "Previous Case Counts| Ebola Hemorrhagic Fever | CDC". Cdc.gov. N.p., 2017. Web. 29 Mar. 2017.
Retrieve date: 03 August 2014
American Scientific Research Journal for Engineering, Technology, and Sciences (ASRJETS) (2017) Volume 37, No 1, pp 1-21
21
[10] Clinaero, Inc. "Ebola Incubation Period". eMedTV: Health Information Brought To Life. N.p., 2017.
Web. 29 Mar. 2017.
[11] "Kermack-Mckendrick Model β From Wolfram Mathworld". Mathworld.wolfram.com. N.p., 2017.
Web. 29 Mar. 2017.
[12] "The SIR Model For Spread Of Disease - Euler's Method For Systems". Mathematical Association of
America. N.p., 2017. Web. 29 Mar. 2017.
[13] Rahman, Prof. Dr. Md. Fazlur. Mathematical Modelling In Biology. 7th ed. ISBN-984-8759-19-0,
2015. Print.
[14] Tsai, Tony. βTony Tsai.β RK4 Method for Solving SIR Model.N.p., n.d. Web. 16 Apr. 2017