Professional Documents
Culture Documents
Forecasting
McGraw-Hill/Irwin Copyright © 2007 by The McGraw-Hill Companies, Inc. All rights reserved.
Learning Objectives
List the elements of a good forecast.
Outline the steps in the forecasting
process.
Describe at least three qualitative
forecasting techniques and the advantages
and disadvantages of each.
Compare and contrast qualitative and
quantitative approaches to forecasting.
3-2
Learning Objectives
Briefly describe averaging techniques,
trend and seasonal techniques, and
regression analysis, and solve typical
problems.
Describe two measures of forecast
accuracy.
Describe two ways of evaluating and
controlling forecasts.
Identify the major factors to consider when
choosing a forecasting technique.
3-3
How the car buyer think ?
He wants it as soon as possible
He does not have to wait any time for delivery
If the dealer is not ready to deliver at once the
buyer look elsewhere
Hence
He usually
Guesses wrong !!!!!!
Gets stuck with cars customers don’t want !!!!!!!
So, how does the dealer know how many cars of each type to stock ?
The answer is :
The dealer doesn’t know for sure, but by analyzing previous buying
patterns, and perhaps making allowances for current conditions, the
dealer can come up with a reasonable approximation of what buyers will
want
3-5
Planning and Uncertainties
The manager is a planner
Uncertainty make planning difficult
Forecasting reduce uncertainty
Forecasting enable managers
to develop more meaningful
plans
so
What is forecasting ?
3-6
What is forecasting ?
FORECASTING IS:
A statement about the future value of a
variable of interest such as demand.
Forecasting is used to make informed
decisions.
Long-range (covering several years) that will
have long-term consequences on a country,
organization, town, city ……. )
Short-range ( several weeks, days,…)
People make and use forecasts all the time
both in their jobs and in everyday life
3-7
Forecasts
Forecasts affect decisions and activities
throughout an organization
Accounting, finance
Human resources
Marketing
MIS
Operations
Product / service design
3-8
Uses of Forecasts
1. Help managers plan the system
( long term )
Types of products to offer
Facilities and equipment to have, to locate
2. Help managers plan the use of the system
( short term )
Planning inventory
Work force level
Planning purchasing and production
Budgeting and scheduling
3-9
Uses of Forecasts
3-10
Features of Forecasts
Assumes causal system
past ==> future
Forecasts rarely perfect because of
randomness
Forecasts more accurate for
groups of items vs. individual items
I see that you will
Forecast accuracy decreases get an A this semester.
3-11
Elements of a Good Forecast
y
el
i m
T
t
le ate ffc
b r t -e
elia c cu
o s
R A C
e
f ul n u s
n g tt e to
ni ri y
a s
M
e W Ea
3-12
Steps in the Forecasting Process
“The forecast”
3-13
Approaches to forecasting
Qualitative:
Consists of subjective inputs
Includes soft information (human factors,
personal opinion)
Quantitative:
Consists of objective inputs
Involve hard data (projection of historical
data, analyzing objectives)
3-14
Types of forecasting techniques
3-15
Judgmental Forecasts
Executive opinions
Sales force opinions
Consumer surveys
Outside opinion
Delphi method
Opinions of managers and staff
Achieves a consensus forecast
3-16
Executive opinions
.Est. * prob probability estimation
131 General
manager
3-17
Delphi method
Session 4 Session 3 Session 2 Session 1
4 4 4.5 5 Expert 1
4 4.5 5 6 Expert 4
3-18
Time Series Forecasts
Plotting the past data and visually
examining the appeared pattern:
Trend - long-term movement in data
Seasonality - short-term regular variations in
data
Cycle – wavelike variations of more than one
year’s duration
Irregular variations - caused by unusual
circumstances
Random variations - caused by chance
3-19
Forecast Variations
Figure 3.1
Irregular
variation
Trend
Cycles
90
89
88
Seasonal variations
3-20
Naive Forecasts
Uh, give me a minute....
We sold 250 wheels last
week.... Now, next week
we should sell....
3-21
Naïve Forecasts
Simple to use
Virtually no cost
Quick and easy to prepare
Data analysis is nonexistent
Easily understandable
Cannot provide high accuracy
Can be a standard for accuracy
3-22
Uses for Naïve Forecasts
Stable time series data
F(t) = A(t-1)
Seasonal variations
F(t) = A(t-n)
Data with trends
F(t) = A(t-1) + (A(t-1) – A(t-2))
3-23
Averaging
White noise and unimportant factors
Minor variation and large variation
Random variation and Real variation
Averaging to smooth data fluctuation
Three types of data changes
3-24
Techniques for Averaging
Moving average
Weighted moving average
Exponential smoothing
3-25
Moving Averages
A technique that averages a number of recent
actual values, updated as new values become
available.
At-1 + At-2….. + At-n
Ft = MAn=
n
Ft = forecast for time period t
MAn = n period moving average
At-1 = actual value in period t-1
n = number of periods ( data points )
3-26
Moving Averages
Example :
Compute a 3-period moving average forecast
given demand for shopping carts for the last
five periods as shown:
5 4 3 2 1 period
6
??? 41 40 43 40 42 demand
3-27
Moving Averages
Example ( cont.) forecast
5 4 3 2 1 period
7 6
??? 41.33 41 40 43 40 42 demand
t=7 actual
6 5 4 3 2 1 period
7
??? 38 41 40 43 40 42 demand
A6 + A 5 + A 4 = 39.67
F7 = MA3=
3
forecast
6 5 4 3 2 1 period
7
39.67 38 41 40 43 40 42 demand
3-28
Moving Averages
#N/A 42 1
#N/A 40 2
#N/A 43 3
-1.6666667 41.666667 40 4
0 41 41 5
-2.3333333 41.333333 39 6
40 7
3-29
Demand Moving Averages
Data
MA3
MA5
Period
3-30
Moving Average
Actual
MA5
47
45
43
41
39
37 MA3
35
1 2 3 4 5 6 7 8 9 10 11 12
3-31
Moving Averages
But
Weighted Moving Averages
3-34
Weighted Moving Averages
Example :
Compute a 4-period weighted moving average forecast given demand
for shopping carts for the last periods values and weights as shown:
5 4 3 2 1 period
6
??? 41 40 43 40 42 demand
F6 = W5*A5 + W4*A4 + W3*A3 + W2*A2
0.4 0.3 0.2 0.1 0
F6 = 0.4(41) + 0.3(40) + 0.2(43) + 0.1(40) = 41 weight
If the actual value of F6 is 39 then
F7 = 0.4(39) + 0.3(41) + 0.2(40) + 0.1(43) = 40.2
3-35
Exponential Smoothing
3-36
Exponential Smoothing
3-38
Picking a Smoothing Constant
Actual
50
.4
D e m a nd
.1
45
40
35
1 2 3 4 5 6 7 8 9 10 11 12
Period
Parabolic
Exponential
Growth
3-40
Techniques for Trend
Linear Trends
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
0 0
3-41
Linear Trend Equation
Ft
Ft = a + bt
0 1 2 3 4 5 t
Ft = Forecast for period t
t = Specified number of time periods
a = Value of Ft at t = 0
b = Slope of the line
3-42
Calculating a and b
n (ty) - t y
b =
n t 2 - ( t) 2
y - b t
a =
n
3-43
Linear Trend Equation Example
t y
2
W eek t S a le s ty
1 1 150 150
2 4 157 314
3 9 162 486
4 16 166 664
5 25 177 885
2
t = 15 t = 55 y = 812 ty = 2 4 9 9
2
( t) = 2 2 5
3-44
Linear Trend Calculation
5 (2499) - 15(812) 12495-12180
b = = = 6.3
5(55) - 225 275 -225
812 - 6.3(15)
a = = 143.5
5
y = 143.5 + 6.3t
3-45
Techniques for Seasonality
Seasonal variations
Regularly repeating movements in series values that can be tied to
recurring events.
It is applied to monthly, weekly, daily and other regularly recurring
patterns in data.
Seasonality is
The amount that the actual values deviate from the average value of
a series.
Seasonal relative
Percentage of average or trend
Centered moving average
A moving average positioned at the center of the data that were
used to compute it.
3-46
MODELS FOR SEASONALITY
ADDITIVE MODEL:
SEASONALITY IS EXPRESSED AS QUANTITY (E.G., 20
UNITS), WHICH IS ADDED TO OR SUBTRACTED FROM
THE SERIES AVERAGE (OR TREND).
MULTIPLICATIVE MODEL:
SEASONALITY IS EXPRESSED AS PERCENTAGE TO BE
MULTIPLIED BY THE AVERAGE OF THE SERIES (OR
TREND), IT IS ALWAYS CALLED SEASONAL RELATIVES
OR SEASONAL INDEX.
3-47
SEASONALITY AND AVERAGE MODELS
Average
Average
3-48
SEASONALITY AND TREND MODELS
Trend
3-49
Seasonal relatives
Seasonal relatives are used in two different
ways in forecasting to:
3-50
Incorporating seasonality into forecast
Example
The manager of a parking lot has
computed daily relatives for the number of
cars per day in the lot. The computations
are repeated here (about three weeks are
shown for illustration). A seven period
centered moving average is used because
there are seven days (seasons) per week
3-54
Calculation of seasonal relatives (cont.)
75 Wed
82 Thur
0.77 71 55 Mon
0.72 72 52 Mon
88 Sat
44 Sun
50 Mon
3-55
Seasonal Relatives (cont.)
The estimated relatives will be:
Monday: (0.77 + 0.72)/2 = .745
Tuesday: (0.84 + 0.89)/2 = .865
Wednesday: (1.03 + 1.06)/2 = 1.045
Thursday: (1.19 + 1.2)/2 = 1.195
Friday: (1.36 + 1.4 + 1.33)/3 = 1.36
Saturday: (1.27 + 1.21)/2 = 1.24
Sunday: (0.51 + 0.56)/2 = 0.535
Note
The sum of the relatives must equal the number of periods
(i.e., 7 in this example). If it is not, you have to multiply by
a correction factor. In this example the sum is 6.985,
therefore you have to multiply each factor by (7/6.985)
3-56
Seasonal Relatives
The number of periods needed in a centered moving
average is equal to the number of “seasons” involved
With monthly data, a 12-period moving average is
needed.
With daily data, a 7-period moving average is needed.
With quarterly data, a 4-period moving average is needed
Note:
When the number of period is even, one additional step is
needed, because the middle of an even period falls
between two periods. The additional step requires taking
two-period moving average of the even-numbered
centered moving average. (See the following example)
3-57
Seasonal relatives
1 4 3 2 1 4 3 2 1 4 3 2 1 Quarter
58 88 84 54 45 71 60 36 28 46 35 18 14 Demand
Note:
The number of seasons is 4 (an even number); therefore, a
4-centered moving average will be taken and then a 2-
centered moving average is calculated to adjust the data.
3-58
Seasonal Relatives
D/MA2 MA2 MA4 Demand (D) Quarter Year
14 1 1
18
28.25
2
1.17 30 35
31.75 3
1.35 34 46 4
36.25
0.71 39.38 28 1 2
42.50
0.79 45.63 36
48.75
2
1.18 50.88 60
53.00 3
1.29 55.25 71
57.50 4
0.74 60.50 45
63.50 1 3
0.82 65.63 54
67.75
2
1.21 69.38 84
71.00 3
88 4
58 1 4
Seasonal relatives
Average relatives
Quarter1 = (0.71+0.74)/2 = 0.725
Quarter2 = (0.79+ 0.82)/2 = 0.805
Quarter3 = (1.17 + 1.18 + 1.21)/3 = 1.187
Quarter4 = (1.35 + 1.29)/2 = 1.32
The sum of these relatives is 4.037. Multiplying
each by 4/4.037 will standardize relatives making
their total equal 4. The resulting relatives are:
Quarter1 = 0.718, Quarter2 = 0.798
Quarter3 = 1.176, Quarter4 = 1.308 3-60
Deseasonalize the data
3-61
Deseasonalize the data
Actual
120 – data (y1)
Sales (millions of dollars)
100 –
80 – Deseasonalized
data (dt)
60 –
40 –
20 –
| | | | | | | | | | | | | | | |
t
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
1998 1999 2000 2001
3-62
Associative Forecasting
Associative models assume that there is a
causational relationship between the
variable of interest and other variables
called predictors
Examples:
• Price of beef and price of chicken
• Crop yields and soil condition
• Crop yields and timing of water application
• Profits and sales
• Price of products and energy cost
Associative Forecasting
Def.
Predictor variables - used to predict
values of variable of interest
Regression - technique for fitting a line
to a set of points
Least squares line - minimizes sum of
squared deviations around the line
3-64
Simple Linear Regression
The simplest and most widely used form of regression
involves a linear relationship between two variables.
The objective in linear regression is to obtain an equation
of a straight line that minimizes the sum of squared
vertical deviations of data points from the line (i.e., the
least squares criterion)
This least squares line has the form:
Y = a + bX
Where:
Y = predicted (dependent or response) variable
X = predictor (independent or explanatory) variable
a = intercept
b = slope (change rate)
3-65
Simple Linear Regression
n (xy) - x y
b =
n x2 - ( x)2
y - b x
a =
n
x2 xy Y X n (xy) - x y
b =
n x2 - ( x)2
49 105 15 7
4 20 10 2
36 78 13 6
16 60 15 4 1.72870662 =b
196 350 25 14
225 405 27 15
y - b x
256 384 24 16 a =
144 240 20 12 n
196 378 27 14 3.5675563 =a
400 880 44 20
225 510 34 15
49 119 17 7 Y = 3.567 + 1.728X
1796 3529 271 132
Simple Linear Regression
If you want to
predict profit at
PROFITS
sales point = 10.3
Profit = 21.365
3-69
PREDICTORES
n (xy) - x y
r =
2 2 2 2
(( ) ( ))
√ n x - ( x)
-1< r < +1
n y - ( y)
•Correlation r close to zero indicate weak linear relationship between two variables
Weakness of regression
Simple linear regression applies only to linear
relationship with one independent variable .
One needs a considerable amount of data to
establish the relationship – in practice, 20 or
more observations.
All observations are weighted equally.
3-73
Simple Regression Analysis
EXAMPLE: Unemp. units
Sales of new houses and period
rate sold
three-month lagged
unemployment are 7.2 20 1
3-75
Simple Regression Analysis
Unemp. rate
Simple Regression Analysis
3-77
Simple Regression Analysis
Y² X² XY x y t
400 51.84 144 7.2 20 1
1681 16 164 4 41 2
289 53.29 124.1 7.3 17 3
1225 30.25 192.5 5.5 35 4
625 46.24 170 6.8 25 5
961 36 186 6 31 6
1444 29.16 205.2 5.4 38 7
2500 12.96 180 3.6 50 8
225 70.56 126 8.4 15 9
361 49 133 7 19 10
196 81 126 9 14 11
b =((11*1750.8)-(305*70.2))/(((11*476.3)-(70.2*70.2)))
b = - 6.91
a = 305 – ( - 6.91 ) / 11
a = 71.85
3-79
Forecast Accuracy
Error - difference between actual value and
predicted value
Mean Absolute Deviation (MAD)
Average absolute error
Mean Squared Error (MSE)
Average of squared error
Mean Absolute Percent Error (MAPE)
Average absolute percent error
3-80
MAD, MSE, and MAPE
Actual forecast
MAD =
n
2
( Actual forecast)
MSE =
n -1
3-81
MAD, MSE and MAPE
MAD
Easy to compute
Weights errors linearly
MSE
Squares error
More weight to large errors
MAPE
Puts errors in perspective
3-82
Example 10
MAD= 2.75
MSE= 10.86
MAPE= 1.28
3-83
Sources of Forecast errors
Model may be inadequate
Irregular variations
Incorrect use of forecasting technique
3-84
Controlling the Forecast
Control chart
A visual tool for monitoring forecast errors
Used to detect non-randomness in errors
Forecasting errors are in control if
All errors are within the control limits
No patterns, such as trends or cycles, are
present
3-85
Control chart
The upper control limit
UCL = 0 + z . MSE
LCL = 0 – z. MSE
Mean = 0
Normal error
due to chance
LCL
Abnormal error
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
period
Tracking Signal
•Tracking signal
The ratio of cumulative error to MAD
Tracking signalt =
Actualt(- )forecastt
MADt
•The tracking signal is computed period by period and it
can be positive or negative, a value of zero would be
ideal; limits of ± 4 or ± 5 are often used for a range of
acceptable values of tracking signal. If a value is out of
the range, then there is a bias in the forecast.
•Bias – Persistent tendency for forecasts to be
Greater or less than actual values.
Tracking signal
The following table shows the calculation of the tracking signal (TS).
Note that all values are within ± 4
TS MADt Cumulative Cumulative e e =A-F Forecast Sales Month
e Error (Error) (F) (A)
-1 4 4 -4 4 -4 47 43 1
0.55 5.5 11 3 7 7 44 51 2
1.4 5 15 7 4 4 50 54 3
2 4 16 8 1 1 50 51 4
0.71 4.2 21 3 5 -5 54 49 5
0.26 3.8 23 1 2 -2 48 46 6
Choosing a Forecasting
Technique
No single technique works in every
situation
Two most important factors
Cost
Accuracy
Other factors include the availability of:
Historical data
Computers
Time needed to gather and analyze the data
Forecast horizon
3-90
Operations Strategy
Forecasts are the basis for many decisions
Work to improve short-term forecasts
Accurate short-term forecasts improve
Profits
Lower inventory levels
Reduce inventory shortages
Improve customer service levels
Enhance forecasting credibility
3-91
Supply Chain Forecasts
Sharing forecasts with supply chain can
Improve forecast quality in the supply chain
Lower costs
Shorter lead times
3-92
Exponential Smoothing
3-93
Linear Trend Equation
3-94
Simple Linear Regression
3-95