Assignment 2
Name Majid Hussain
Reg. # 18 – IE – 10
Question:
Use WHO data from last 8 months relevant to virus cases of the world, forecast cases for 9th
month. Validate your forecast and conclude which forecast method gives best results. Also figure
out relation, if any, of population, poverty and GDP of countries with virus cases, use fifteen
countries for this relation.
Solution:
The WHO monthly data of corona virus cases is collected from the following website link from
February 2020 to September 2020.
[Link]
The Table Below gives us the value of confirmed cases in the world for the given months:
Month Number of Confirmed Cases
February 81408
March 1110179
April 2363006
May 2985204
June 4644455
July 7364017
August 8154518
September 8544782
The cases for the month of October is 13037316. Now we shall see which method of forecasting
gives us the best results.
We shall use the POM QM for Windows software to calculate the forecasting results.
We shall use the following Time Series Analysis Method as given below:
1. Naïve Method
2. A 3-month Moving Averages
3. A 6-month Weighted moving averages
4. Exponential Smoothing (with smoothing constant α = 0.99)
5. Trend adjusted exponential smoothing (with α = 0.99 and β = 0.99)
Naïve Method:
When we enter the data by selecting the Naïve Method from the dropdown menu the window
screen looks like as shown below:
The forecasting of confirmed cases for the month of October is given below:
In the forecasting results window, we can see in the figure shown above that it forecasts the
number of cases in October is 8544782 with cumulative forecast error value which is 8463374,
MAD (Mean Absolute Deviation) which is 1209053, MSE (Mean Squared Error) which is
1991629000000, Standard Error which is 1669815 and MAPE (Mean Absolute Percent Error)
which is 36.21%.
As the original number of cases in October is 13037316. So, the forecast error is:
Forecast error = 4492534
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
A 3-month moving Average:
When we enter the data by selecting the moving average method from the dropdown menu and
click on the solve button the results window is shown below:
In the forecasting results window, we can see in the figure shown above that it forecasts the
number of cases in October is 8021106 with cumulative forecast error value which is 13305540,
MAD (Mean Absolute Deviation) which is 2661108, MSE (Mean Squared Error) which is
7801239000000, Standard Error which is 3605838 and MAPE (Mean Absolute Percent Error)
which is 45.76%.
As the original number of cases in October is 13037316. So, the forecast error is:
Forecast error = 5016210
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
A 6-month weighted moving average:
When we enter the data by selecting the moving average method from the dropdown menu and
click on the solve button. The weightage table is given below:
Month Weightage
February 0.1
March 0.2
April 0.3
May 0.4
June 0.5
July 0.6
The forecasting results are shown in below figure:
In the forecasting results window, we can see in the figure shown above that it forecasts the
number of cases in October is 4506084 with cumulative forecast error value which is 11540600,
MAD (Mean Absolute Deviation) which is 5770299, MSE (Mean Squared Error) which is
33478750000000, Standard Error which is NA and MAPE (Mean Absolute Percent Error) which
is 69.27%.
As the original number of cases in October is 13037316. So, the forecast error is:
Forecast error = 8531232
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
Exponential Smoothing:
As we know that 99% of the number of cases that will be registered in the month of October
depends on the number of number of cases in previous months because Corona Virus is
Pandemic disease which spread by social contact between the people also. That’s why we are
going to use a value of 0.99 for the smoothing constant.
So, we enter the data by selecting the Exponential Smoothing method from the dropdown menu
and click on the solve button. The value of α is 0.99 and the forecasts results are shown in the
figure below:
In the forecasting results window, we can see in the figure shown above that it forecasts the
number of cases in October is 8540798 with cumulative forecast error value which is 8544838,
MAD (Mean Absolute Deviation) which is 1068105, MSE (Mean Squared Error) which is
1768200000000, Standard Error which is 1535448 and MAPE (Mean Absolute Percent Error)
which is 31.89%.
As the original number of cases in October is 13037316. So, the forecast error is:
Forecast error = 4496518
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
Trend adjusted exponential smoothing:
Again, as we know that 99% of the number of cases that will be registered in the month of
October depends on the number of number of cases in previous months because Corona Virus is
Pandemic disease which spread by social contact between the people also. That’s why we are
going to use a value of 0.99 for the smoothing constant.
we enter the data by selecting the Trend Adjusted Exponential Smoothing method from the
dropdown menu and click on the solve button. The value of α is 0.99 and β is also 0.99 and the
forecasts results are shown in the figure below:
In the forecasting results window, we can see in the figure shown above that it forecasts the
number of cases in October is 8929981 with cumulative forecast error value which is 388243.8,
MAD (Mean Absolute Deviation) which is 914869.6, MSE (Mean Squared Error) which is
1087545000000, Standard Error which is 1233922 and MAPE (Mean Absolute Percent Error)
which is 27.18%.
As the original number of cases in October is 13037316. So, the forecast error is:
Forecast error = 4107335
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
Conclusion:
After comparing the results of all the forecasting methods forecasted error, we came to know that
Trend Adjusted Exponential Smoothing is the forecast method that gives us best results.
Relation of Population with Corona Viruses cases in a Country:
In order to find the relation of Corona Viruses cases with the Population, we shall use Linear
Regression/ Least Squares method of Time Series Analysis. The cases in 15 different countries
of the world and their population is given below in the table:
Country Virus Cases Population
United States of America 9108353 328200000
India 8267623 1353000000
Brazil 5545705 209500000
Russian Federation 1655034 144500000
France 1381098 66990000
Spain 1185678 46940000
Argentina 1173533 44490000
Colombia 1083321 49650000
The United Kingdom 1034918 66650000
Mexico 929392 126200000
Peru 904911 31990000
South Africa 726823 57780000
Italy 709335 60360000
Iran 628780 81800000
Pakistan 335093 212200000
When we enter these values in POM QM for windows and click on the Solve button, we got the
following results as shown in figure below:
In the forecasting results window, we can see in the figure shown above that it gives us the value
of correlation coefficient 0.72 with cumulative forecast error value which is -1.75, MAD (Mean
Absolute Deviation) which is 1209961, MSE (Mean Squared Error) which is 3672594000000,
Standard Error which is 2058545 and MAPE (Mean Absolute Percent Error) which is 95.71%.
The relation between Population and Corona Cases is given by following equation:
Cases=1115799+ 0.1× Population
The Correlation Coefficient value is 0.72 and Coefficient of Determination value is 0.52 for this
model. As we know that correlation coefficient (r) is given as a measure of linear association
between the two variables. It means 72 percent values of the two variables in our data are
linearly associated which means increasing the value of one variable will increase the value of
another variable.
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
Conclusion:
From the above discussion, it is clear that the number of Virus cases in a country are dependent
on its population or in other words 72% chances exist that if a country has a large population, it
will also have a large number of Virus Cases.
Relation of Poverty rate with Corona Viruses cases in a Country:
In order to find the relation of Corona Viruses cases with the Poverty rate, we shall use Linear
Regression/ Least Squares method of Time Series Analysis. The cases in 15 different countries
of the world and their poverty rate is given below in the table:
Country Virus Cases Poverty Rate
United States of America 9108353 10.5
India 8267623 68.8
Brazil 5545705 20.4
Russian Federation 1655034 13.1
France 1381098 14
Spain 1185678 21.6
Argentina 1173533 9.60
Colombia 1083321 27.80
The United Kingdom 1034918 23.2
Mexico 929392 6.6
Peru 904911 20.5
South Africa 726823 49.2
Italy 709335 7.7
Iran 628780 10.9
Pakistan 335093 24.3
When we enter these values in POM QM for windows and click on the Solve button, we got the
following results as shown in figure below:
In the forecasting results window, we can see in the figure shown above that it gives us the value
of correlation coefficient 0.34 with cumulative forecast error value which is 1.25, MAD (Mean
Absolute Deviation) which is 1866342, MSE (Mean Squared Error) which is 6752134000000,
Standard Error which is 2791223 and MAPE (Mean Absolute Percent Error) which is 144.81%.
The relation between Population and Corona Cases is given by following equation:
Cases=1032811 +58432.13× P overty Rate
The Correlation Coefficient value is 0.34 and Coefficient of Determination value is 0.12 for this
model. As we know that correlation coefficient (r) is given as a measure of linear association
between the two variables. It means 34 percent values of the two variables in our data are
linearly associated which means increasing the value of one variable will increase the value of
another variable.
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
Conclusion:
From the above discussion, it is clear that the number of Virus cases in a country are dependent
to some extent on its poverty rate or in other words 34% chances exist that if a country has a
large poverty rate, it will also have a large number of Virus Cases.
Relation of GDP with Corona Viruses cases in a Country:
In order to find the relation of Corona Viruses cases with the GDP, we shall use Linear
Regression/ Least Squares method of Time Series Analysis. The cases in 15 different countries
of the world and their poverty rate is given below in the table:
Country Virus Cases GDP(Trillion USD)
United States of America 9108353 20.54
India 8267623 2.719
Brazil 5545705 1.869
Russian Federation 1655034 1.658
France 1381098 2.778
Spain 1185678 1.419
Argentina 1173533 0.5199
Colombia 1083321 0.331
The United Kingdom 1034918 2.855
Mexico 929392 1.221
Peru 904911 0.222
South Africa 726823 0.3683
Italy 709335 2.084
Iran 628780 0.454
Pakistan 335093 0.3146
When we enter these values in POM QM for windows and click on the Solve button, we got the
following results as shown in figure below:
In the forecasting results window, we can see in the figure shown above that it gives us the value
of correlation coefficient 0.71 with cumulative forecast error value which is -2, MAD (Mean
Absolute Deviation) which is 1261042, MSE (Mean Squared Error) which is 3747816000000,
Standard Error which is 2079520 and MAPE (Mean Absolute Percent Error) which is 88.49%.
The relation between Population and Corona Cases is given by following equation:
Cases=1248040+ 405282.3 ×GDP
The Correlation Coefficient value is 0.71 and Coefficient of Determination value is 0.51 for this
model. As we know that correlation coefficient (r) is given as a measure of linear association
between the two variables. It means 71 percent values of the two variables in our data are
linearly associated which means increasing the value of one variable will increase the value of
another variable.
A graph with two curves in which one curve is showing the Actual Demand and second curve is
showing the Forecasted Demand is also shown below in figure:
Conclusion:
From the above discussion, it is clear that the number of Virus cases in a country are dependent
on its GDP or in other words 71% chances exist that if a country has a large GDP, it will also
have a large number of Virus Cases.
