Simulation and Forecasting Solutions

Question 1

(a) The concept of the Monte Carlo simulation is experimentation on the probabilistic
elements through random sampling.

The major steps used in the Monte Carlo simulation process are as follows:

1. Set up a probability distribution and the cumulative probability distribution of the

past performance.
2. Establish interval of random numbers.
3. Generate random numbers using the random number tables.
4. Perform simulation by mapping random number with input values.

(b)(i) The Monte Carlo Simulation

Number Probability Random Random Simulated

of Cars of Interval Number Demand
Month (From Table) (No. of cars)
Sold Demand Range
1 5 0.16 00 - 15 62 10
2 8 0.25 16 – 40 32 8
3 2 0.06 41 – 46 71 10
4 10 0.31 47 – 77 94 5
5 2 0.06 78 – 83 04 5
6 5 0.16 84 - 99 97 5
Total 0 0.00 0

(b)(ii) Simulated monthly demand = 43/6 =7.17 or 7 cars

(b(iii) Expected Monthly Demand

= (Probability of Demand)  (Demand per Month)
= 0.16(5) + 0.25(8) + 0.06(2) + 0.31(10) + 0.06((2) + 0.16(5)
= 6.94 or 7 cars

Question 2

(a) i)
Sales Interval

1|Page
 0 .00 but less than .08
1 .08 but less than .20
2 .20 but less than .48
3 .48 but less than .72
4 .72 but less than .86
5 .86 but less than .96
6 .96 but less than 1.00

ii) 2, 5, 2, 3, 2, 4, 2, 1, 1, 2

iii) Total Sales = 24 units

(b)
1. It can be used for a wide variety of practical problems.
2. The simulation approach is relatively easy to explain and understand. As a result,
management confidence is increased and acceptance of the results is more easily
obtained.
3. Spreadsheet packages now provide another alternative for model implementation,
and third-party vendors have developed add-ins that expand the capabilities of
the spreadsheet packages.
4. Computer software developers have produced simulation packages that make it
easier to develop and implement simulation models for more complex problems.

Question 3

(a) The following table can be used to simulate a win for MDIS University

Game Interval for MDIS University Win

1 .00 but less than .60
2 .00 but less than .55
3 .00 but less than .48
4 .00 but less than .45
5 .00 but less than .48
6 .00 but less than .55
7 .00 but less than .50

(b) Using the random numbers, MDIS University wins games 1 and 2, loses game 3, wins
game 4, loses game 5 and wins game 6. Thus, MDIS University wins the 6-game
Intervarsity Series 4 games to 2 games.

(c) - Simulation provides insight into the problem solution when other management
science methods fail.
- Simulation enables us to project the performance of an existing system under a
proposed set of modifications without disrupting current system performance.
- Simulation can be performed over any time horizon.
- Simulation models assist in the design of proposed systems by providing a
convenient experimental laboratory for conducting “what if” analyses.

2|Page
(d) It is important to realize forecasting is not an exact science. No matter which
technique is selected, the goal is the same – to develop a prediction regarding the future that
is more accurate than simply guessing.

Possible forecast errors include data collection, data entry, arithmetical errors, etc. However,
even if it were true that factors that have affected patterns of economic activity in the past and
present will continue to do so in a similar manner in the future, then forecasting would
certainly be an appropriate tool to managerial control of present activities.

However, most forecasting techniques are overly naïve and mechanical. A mathematical
model based on the past should not be used to mechanically extrapolate trends into the future
without considering personal judgments, business experiences or changing technologies,
habits and needs. Thus, econometricians have been concerned with developing more highly
sophisticated computerized models of economic activity for forecasting purposes.

Question 4

(a)
Type of Cost Amount ($’000) Interval
14 0.00 to 0.25
16 More than 0.25to 0.60
Facilities
18 More than 0.60 to 0.90
20 More than 0.90 to 1
3 0.00 to 0.10
5 More than 0.10 to 0.35
Equipment
7 More than 0.35 to 0.80
9 More than 0.80 to 1.00
24 0.00 to 0.05
26 More than 0.05 to 0.30
Installers
28 More than 0.30 to 0.70
30 More than 0.70 to 1.00

(b)
Type of cost Random Number Amount ($’000)
Facilities 0.9167 20
Equipment 0.7619 7
Installers 0.2681 26
Total Cost 53

Profit = $56,000 - $53,000 = $3,000 per chair

(c) Simulation will provide a distribution of the profit per unit values. Calculating the
percentage of simulation trials providing a profit less than $5000 per chair would provide an
estimate of the probability that the profit per unit will be unacceptably low.

3|Page
(d) - Able to solve wide variety of business problems
- Easy to explain and understand
- Spreadsheet packages readily available
- Computer software available

Question 5

(a) (i) to (iii)

Month Sales 3 mth Moving Absolute Exponential Absolute

Average Deviation Smoothing Deviation
January 440
February 520
March 640 520.00
April 760 533.33 226.67 556.00 204.00
May 540 640.00 100.00 617.20 77.20
June 480 646.67 166.67 594.04 114.04
July 560 593.33 33.33 559.83 0.17
August 530 526.67 3.33 559.88 29.88
September 620 523.33 96.67 550.92 69.08
Mean Absolute Deviation 104.44 82.40

Based upon MAD, the exponential smoothing model appears to be the better choice.

b) Of these two choices, the weighted moving average is the most accurate, since
specific weights can be placed in accordance with their importance. The other method make
assumptions, such as an average. The weighted average may be modified to any form. If a
long time span is taken, however, the weighted average can be cumbersome to use. In
addition, as time periods pass, the user most likely would like to change the weights. This
would add to the difficulty of using the techniques for a large number of applications such as
forecasting demand for inventory items.

Question 6

(a)
(i) FORECASTING WITH EXPONENTIAL SMOOTHING
THE SMOOTHING CONSTANT IS α= 0.2

4|Page
 Time Period Actual Value Forecast Forecast Error
1
2 46 58.00 -12.00
3 55 55.60 -0.60
4 39 55.48 -16.48
5 42 52.18 -10.18
6 63 50.15 12.85
7 54 52.72 1.28
8 55 52.97 2.03
9 61 53.38 7.62
10 52 54.90 -2.90

THE MEAN SQUARE ERROR = [(-12.00)2+(-0.60)2+(-16.48)2+

( 10.18)2+(12.85)2+(1.28)2+(2.03)2+( 7.62)2+(-2.90)2]/9= 756.939/9 = 84.10

THE FORECAST FOR PERIOD 11 54.32

(ii) FORECASTING WITH EXPONENTIAL SMOOTHING

THE SMOOTHING CONSTANT IS α= 0.8

Time Period Actual Value Forecast Forecast Error

1
2 46 58.00 -12.00
3 55 48.40 6.60
4 39 53.68 -14.68
5 42 41.94 0.06
6 63 41.99 21.01
7 54 58.80 - 4.80
8 55 54.96 0.04
9 61 54.99 6.01
10 52 59.80 -7.80

THE MEAN SQUARE ERROR = [(-12.00)2+(6.60)2+(-14.68)2+( 0.06)2+(21.01)2+(-

4.80)2+(0.04)2+( 6.01)2+(-7.80)2]/9= 964.4878/9 =107.17

THE FORECAST FOR PERIOD 11 53.56

Based on lower MSE, smoothing with α = .2 provides a better model.

(b) The objective of smoothing methods is to “smooth out” random fluctuations

caused by the irregular component of a time series.

Such methods are appropriate for a stable time series (i.e. a time series that

5|Page
 does not show significant trend, cyclical or seasonal effects) because they
adapt well to changes in the level of the time series.

Smoothing methods are easy to use and provides a reasonably good degree of
accuracy for short-range forecast such as a forecast of the next time period.

One of the methods, exponential smoothing, has minimal data requirements and thus is a
good method to use when forecasts are required for large number of items.

Question 7

(a)

Flights to Destination A Flights to Destination B

Make Reservations Rel. Freq. Interval Make Reservations Rel. Freq. Interval
Less than 1 week .25 0.00 to 0.25 Less than 1 week .12 0.00 to 0.12
in advance in advance
1 week to 2 months .35 More than 0.25 1 week to 2 months .35 More than
in advance to 0.60 in advance 0.12 to 0.47
Over 2 months .40 More than 0.60 2 months to 6 months .40 More than
in advance to 1.00 in advance 0.47 to 0.87
Over 6 months .13 More than
In advance 0.87 to 1.00

Flight Destinations Rel. Freq. Interval

A .65 0.00 to 0.65
B .35 More than 0.65
to 1.00

(b)
Random
Observation RN Destination Booking
No.
1 .632 A .715 Over 2 months
2 .998 B .671 2 to 6 months
3 .774 B .021 Less than 1 week

(c) To use the same simulation model to determine the average profits per customer, we
will need to do the following steps:

(i) First calculate the average profits per customer for each destination.
Cost per passenger = Total cost per aircraft flight / seating capacity
Cost per passenger (Destination A) = 11,250/150 = $75/customer.
Cost per passenger (Destination B) = 7,500/150 = $50/customer.

(ii) As the time from a reservation to actual flight is not relevant for this profit
scenario, we can ignore the first and second tables. We only need to use the
third table showing 0.65 for Destination A and 0.35 for Destination B. If
Destination A, profit =$75. If Destination B, profit = $50.

6|Page
 (iii) We then fixed the total number of simulation runs, for e.g. 1,000 runs or
observations. For this, we can use 1,000 random numbers from a random
number generator or random number table.

(iv) Using the same 6 random numbers to start the simulation as an example, we
then tabulate the results for each observation until the 1,000th observation.

Observation RN Destination Profit

1 .632 A 75
2 .715 B 50
3 .998 B 50
4 .671 B 50
5 .774 B 50
6 .021 A 75
..
1000

(v) Summing up the profits for all 1,000 observations, we then compute the
average profit per customer as follows:
Average profit per customer = total profits for 1000 observations / 1000
observations.

Question 8

(a) FORECASTING WITH MOVING AVERAGES

THE MOVING AVERAGE USES 3 TIME PERIODS

Time Period Actual Forecast Forecast Forecast Absolute

Value Error Square Error Forecast Error

2
3
4 265 278.33 -13.33 177.69 13.33
5 314 254.00 60.00 3600.00 60.00
6 323 285.67 37.33 1393.53 37.33
7 299 300.67 -1.67 2.79 1.67
8 259 312.00 -53.00 2809.00 53.00
9 287 293.67 -6.67 44.49 6.67
10 302 281.67 20.33 413.31 20.33
11 282.67

THE FORECAST FOR PERIOD 11 =(312+293.67+281.67)/3 =848/3= 282.67

(b) See Forecast Square error column and Absolute Forecast error in above table

Mean Square error = Sum of Squares error / no. of forecasts

= (177.69+3600.00+1393.53+2.79+2809.00+44.49+413.31) /7

7|Page
 = 8440.81 / 7
= 1205.83

Mean Absolute Deviation = Sum of absolute forecast error / no. of forecasts

= (13.3+ 60.00+37.33+1.67+53.00+6.67+20.33) /7
= 192.33/ 7
= 27.48

Compared to mean square error, the mean absolute deviation measure is less sensitive
to individual large forecast errors than the mean squared error measure

(c) It is important that students realized forecasting is not an exact science. No matter
which technique is selected, the goal is the same – to develop a prediction regarding the
future that is more accurate than simply guessing.

Possible errors include inaccurate counting, data entry, arithmetic calculations, changes
in drinking patterns due to unusual changes in weather, casual factors like sales
campaigns. If manufacturer promotes the drink, the sales goes up. If competitor
promotes a competing drink, sales may go down. There may also be Stock out should
there be a sudden surge in sales or replenishment of vending machines is not reliable.

Most forecasting techniques are overly naive and mechanical. A mathematical model
based on the past should not be used to mechanically extrapolate trends into the future
without considering personal judgments, business experiences or changing
technologies, habits and needs. Thus, econometricians have been concerned with
developing more highly sophisticated computerized models of economic activity for
forecasting purposes.

Question 9

(a)
Minutes Relative Frequency DT x Rel.Freq
3 .14 0.42
4 .30 1.2
5 .27 1.35
6 .18 1.08
7 .11 0.77
Average drying time = 0.42+1.2+1.35+1.08+0.77=4.82min

(b)
(i)
Minutes Relative Frequency Random No. Interval
3 .14 0.00 but less than 0.14
4 .30 0.14 but less than 0.44

8|Page
 5 .27 0.44 but less than 0.71
6 .18 0.71 but less than 0.89
7 .11 0.89 but less than 1

(ii)

Random Number Drying Time

.33 4
.09 3
.19 4
.81 6
.12 3
.88 6
.53 5
.95 7
.77 6
.61 5
.91 7
.47 5

Average drying time = Total drying times/12 observations

= (4+3+4+6+3+6+5+7+6+5+7+5)/12
= 61/12
= 5.0833 minutes

(c) The three possible reasons for the differences are:

(1) The actual experiment to determine the drying times frequencies may not be
comprehensive enough. It only captures drying times from 3 minutes up to 7
minutes. There may be drying times of 1 minute, 2 minutes or occurrences of
8 minutes, 9 minutes or more.

(2) Insufficient simulation runs (or observations). As the number of simulation

runs increases, the average drying rate from the simulation will tend towards
the mean drying time.

(3) Rounding errors (to the nearest minute) has occurred as the actual drying times
from the actual experiment will most likely be in decimals (e.g. 3.5 minutes)
rather than whole numbers.

Question 10

(a)
5-Day Moving Forecast
Day Time-Series Average Forecast Error (Error)2

9|Page
 Value
1 14.45
2 15.75
3 16.45
4 17.40
5 17.32
6 15.96 16.27 -0.31 0.10
7 16.45 16.58 -0.13 0.02
8 15.60 16.72 -1.12 1.25
9 15.09 16.55 -1.46 2.13
10 16.42 16.08 0.34 0.12
11 16.21 15.90 0.31 0.10
12 15.22 15.95 -0.73 0.53

Note: MSE = 4.25/7 = 0.61

Forecast for September 4 is (15.60 + 15.09 + 16.42 + 16.21 + 15.22)/5 = 15.71
(3 marks)

(b) The exponential smoothing forecasts for days 2-12 are 14.45, 15.36, 16.12, 17.02,
17.23, 16.34, 16.42, 15.85, 15.32, 16.09 and 16.17

Note: MSE = 9.57/11 = 0.87

Forecast for September 4 is 0.7(15.22) + 0.3(16.17) = 15.51

(c)
Method MSE
Moving Averages 0.60
Exponential Smoothing 0.87

Moving Averages is the better of the two approaches because it has the smallest MSE.

Question 11

(a)
Payment Interval
$0 .00 but less than .83
500 .83 but less than .89
1,000 .89 but less than .94
2,000 .94 but less than .96
5,000 .96 but less than .98
8,000 .98 but less than .99
10,000 .99 but less than 1.00

(b)
Random Number Probability Payment
69393 0.69393 0
92785 0.92785 1000

10 | P a g e
 49902 0.49902 0
58447 0.58447 0
42048 0.42048 0
30378 0.30378 0
87618 0.87618 500
26933 0.26933 0
40640 0.40640 0
16281 0.16281 0

2 claims made
Total amount paid to the policyholders was $1500

(c) Students are expected to elaborate and explain how simulation can be useful to the
following list. However, they are allowed to discuss areas outside the listing provided
that good elaboration and explanation are provided.

 New Product Development

 Airline Overbooking
 Inventory Policy
 Traffic Flow
 Waiting Lines

(d) Advantages:

1. Simulation provides insight into the problem solution when other management
science methods fail.

2. Simulation enables us to project the performance of an existing system under a

proposed set of modifications without disrupting current system performance.
Such performance may be analysed over any time horizon.

3. Simulation models assist in the design of proposed systems by providing a

convenient experimental laboratory for conducting “what-if” analyses.

Disadvantages:

1. Simulation models are generally time consuming and expensive to develop.

2. Simulation models provide only an estimate of a model’s true parameter

values.

3. There is no guarantee that the policy shown to be optimal by the simulation is

in fact, optimal.

Question 12

(a)
Time-Series α = 0.1 α = 0.1 α = 0.2 α = 0.2

11 | P a g e
 Season Value Forecast (Error)2 Forecast (Error)2
2002 29.6
2003 28.7 29.60 0.81 29.60 0.81
2004 26.8 29.51 7.34 29.42 6.86
2005 29.7 29.24 0.21 28.90 0.64
2006 31.1 29.29 3.29 29.06 4.16
2007 31.4 29.47 3.73 29.47 3.72
Totals: 15.40 16.19

Smoothing Constant MSE

0.1 15.25/5 = 3.05
0.2 16.19/5 = 3.24

A smoothing constant of 0.1 is better.

(b) Using α = 0.1, forecast for 2008 season = 0.1(31.4) + 0.9(29.47) = 29.66

(c) The trend component accounts for the gradual shifting of the time series over a long
period of time.

Any regular pattern of sequences of values above and below the trend line is
attributable to the cyclical component of the series.

The seasonal component of the series accounts for regular patterns of variability within
certain time periods, such as over a year.

The irregular component of the series is caused by short-term, unanticipated and non-
recurring factors that affect the values of the time series. One cannot attempt to
predict its impact on the time series in advance.

Question 13

(i)
Sales Interval
0 .00 but less than .08
1 .08 but less than .20
2 .20 but less than .48
3 .48 but less than .72
4 .72 but less than .86
5 .86 but less than .96
6 .96 but less than 1.00

(ii)
Random Numbers 0.75947 0.66466 0.06232 0.10913 0.75336

12 | P a g e
 Sales (units) 4 3 0 1 4

(iii) Total books sold in five day = 12

Question 14
(a)
Type of Cost Amount ($’000) Interval
14 0.00 to 0.25
16 More than 0.25to 0.60
Facilities
18 More than 0.60 to 0.90
20 More than 0.90 to 1
3 0.00 to 0.10
5 More than 0.10 to 0.35
Equipment
7 More than 0.35 to 0.80
9 More than 0.80 to 1.00
24 0.00 to 0.05
26 More than 0.05 to 0.30
Lecturer
28 More than 0.30 to 0.70
30 More than 0.70 to 1.00

(b)
Type of cost Random Number Amount ($’000)
Facilities 0.9617 20
Equipment 0.4503 7
Lecturer 0.1862 26
Total Cost 53

Profit = $56,000 - $53,000 = $3,000 per students

(c) Simulation will provide a distribution of the profit per unit values. Calculating the
percentage of simulation trials providing a profit less than $5000 per student would
provide an estimate of the probability the profit per unit will be unacceptably low.

(d) - Able to solve wide variety of business problems

- Easy to explain and understand
- Spreadsheet packages readily available
- Computer software available

Question 15

(a)

13 | P a g e
 Cumulativ
Activity Time Random
Activity Probability e
(Months) Number
Probability
10 0.15 0.15
12 0.25 0.40 0.3726
A
14 0.35 0.75
16 0.25 1.00

8 0.20 0.20
B 11 0.25 0.45
12 0.55 1.00 0.5839

5 0.20 0.20
6 0.05 0.25
C 7 0.10 0.35 0.3496
8 0.20 0.55
9 0.45 1.00

13 0.20 0.20 0.1075

D
16 0.80 1.00

Simulated completion time = 12 + 12 + 7 + 13 = 44 months

(b) Simulation will provide a distribution of project completion time values. Calculating
the percentage of simulation trials with completion times of 35 weeks or less can be
used to estimate the probability of meeting the completion time target of 35 weeks.

(c) Using simulation, we can ask what-if questions and project how the real system will
behave. Although simulation does not guarantee optimality, it will usually provide
near-optimal solutions. In addition, simulation models often warn against poor
decision strategies by projecting disastrous outcomes such as system failures, large
financial losses, and so on.

For disadvantages, the process of developing, verifying and validating can be time-
consuming and expensive. In addition, each simulation run provides only a sample of
how the real system will operate. As such, the summary of the simulation data
provides only estimates or approximations about the real system. Consequently,
simulation does not guarantee an optimal solution.

Question 16

14 | P a g e
(a) (i) Starting values can be simply an average of the early periods, or a guess. If the
starting value is taken some period back (as opposed to starting to use the
equations on very recent data) the equation will have a chance to adjust as it is
carried forward to today.

(ii) Multiplicative effects are the most common. Essentially all business sales are
multiplicative. For example, in auto sales, as growth in total sales increases,
the seasonal variation increases. Students may provide any suitable examples.

(iii) The main disadvantage is the mathematical computation. There is very little
time involved for a computerized analysis, but hand computation is time
consuming and prone to error.

(b) (i) FOctober = (410 + 310 + 410)/3 = 376.67

(ii) FOctober = 0.25(410) + 0.30(310) + 0.45(410) = 380

(iii)
Month Sales ($’000) Exponential Smoothing
January 110 130.00
February 310 125.00
March 510 171.25
April 710 255.94
May 610 369.45
June 810 429.59
July 410 524.69
August 310 496.02
September 410 449.51
October 439.63

Question 17

(a)
Year Inflation Rate (%) Moving Average Absolute Deviation
2001 1  
2002 -0.4  
2003 0.5  
2004 1.7 0.3667 1.7778
2005 0.5 0.6000 0.0100
2006 1 0.9000 0.0100
2007 2.1 1.0667 1.0678
2008 6.6 1.2000 29.1600
2009 0.6 3.2333 6.9344
2010 2.8 3.1000 0.0900

15 | P a g e
 2011   3.3
MSE 5.6

b)
Year Inflation Rate (%) Exponential Smoothing Absolute Deviation
2001 1    
2002 -0.4 1.0000  
2003 0.5 0.0200 0.2304
2004 1.7 0.3560 1.8063
2005 0.5 1.2968 0.6349
2006 1 0.7390 0.0681
2007 2.1 0.9217 1.3884
2008 6.6 1.7465 23.5563
2009 0.6 5.1440 20.6475
2010 2.8 1.9632 0.7003
2011   2.5  
    MSE 6.1

c) Of these two choices, the moving average is more accurate, since the MSE is smaller.

d) Advantages:

1. Simulation provides insight into the problem solution when other management
science methods fail.

2. Simulation enables us to project the performance of an existing system under a

proposed set of modifications without disrupting current system performance.
Such performance may be analysed over any time horizon.

3. Simulation models assist in the design of proposed systems by providing a

convenient experimental laboratory for conducting “what-if” analyses.

Disadvantages:

1. Simulation models are generally time consuming and expensive to develop.

2. Simulation models provide only an estimate of a model’s true parameter

values.

3. There is no guarantee that the policy shown to be optimal by the simulation is

in fact, optimal.

Question 18

(a) Generate random numbers for service times by each customer based on the
following table ranges:

16 | P a g e
 Service Time (minutes) Probability Range
2 0.24 0 to 0.24
3 0.20 0.24 to 0.44
4 0.15 0.44 to 0.59
5 0.14 0.59 to 0.73
6 0.12 0.73 to 0.85
7 0.08 0.85 to 0.93
8 0.05 0.93 to 0.98
9 0.02 0.98 to 1.00

Generate inter-arrival times between consecutive customers according to:

T = 1.5 LN(X)
where LN is the natural logarithm.

Other points:
- Sum of arrival time and service time for first customer to obtain completion time for
first customer.
- If the completion time is less than the arrival time of the second customer then the
second customer do not have to wait.
- The arrival time of the second customer is given by the sum of arrival time of first
customer and inter-arrival time between the first and second customer.
- The comparison of completion time of preceding customer and arrival times of the
subsequent customers are made across all the runs and the number of cases with
completion time higher than the arrival times of subsequent customers are cases
which customers have to wait.
- Proportion of cases which customers have to wait against the total number of
simulation runs will provide the estimate of the probability that customers have to
wait in queue.
- Good to evaluate cases after simulation have reached a steady state.

(b) Practical examples may differ.

- Using simulation, we can ask what-if questions and project how the real system will
behave.
- Although simulation does not guarantee optimality, it will usually provide near-
optimal solutions.
- In addition, simulation models often warn against poor decision strategies by
projecting disastrous outcomes such as system failures, large financial losses, and so
on.
- Other relevant advantages can be accepted.

Question 19

(a) With α = 0.2, F13 = 3,213

With α = 0.8, F13 = 3,579

17 | P a g e
  Month 11 has a highest volume compared to previous months, this volume
reduced in month 12.
 The difference in forecasts in month 13 is smaller than month 12.
 Higher α value puts more weight on the latest observations.
 Other similar observations can be accepted.

(b)
Month 3-months moving average
11 3137
12 3567
13 3700

(c) Possible criteria are Mean Squared Errors and Mean Absolute Deviations between the
forecasts and the observations. Any valid discussions related to these criteria can be accepted

18 | P a g e