Demand Forecasting

Dr.
Weria Khaksar
Email: weria@uniten.edu.my
Room No. BN-03-08
Accounting:
product/process
cost estimates,
ForecastsNew
are
a basic input
in the
cash management,
decision
making
processes
Finance: Equipment needs, Amount of funding,
because
they
provide
Human Resource: Hiring and Layoff activities,
information
on
FUTURE
Marketing: Pricing and Promotion,
e-business
strategies,
global competition strategies,
DEMAND.
Operations: Scheduling, capacity planning, work
Why?
assignments, inventory
planning, project
management,
Because
the primary goal of the
Product/Service Design: Revision of current
production
planners
is
to
match
features, design of new products or services,
1. They assume that the same system that

existed in the past, will continue in the
future.
2. Forecasts are not perfect and actual
results usually differ from predicted
values (ERROR).
3. Forecast of groups of items tend to be
more accurate than individuals.
4. Forecast accuracy decreases as the
time period covered by the forecast (Time
The forecast should be:
1. Timely
2. Accurate
3. Reliable
4. Meaningful
5. Based
on
a
simple
understandable method
6. Cost-effective
and
1. Determine the purpose of the forecast.

2. Establish a time horizon.
3. Select a forecasting technique.
4. Obtain, clean, and analyze appropriate data.
5. Make the forecast.
6. Monitor the forecast.
Accuracy of a forecast is minimizing the

forecast error.
Accuracy if one of the factors for choosing a
forecasting technique.
Forecast ERROR is the difference between the

actual and predicted values:
=
Positive errors happen when forecast is too
low.
Negative errors happen when forecast is too
high.
Three commonly used measures for summarizing

historical errors are:
1. Mean Absolute Deviation (MAD): Average absolute
error
=
2. Mean Squared Error (MSE): Average of squared errors

( )
=
=
3. Mean absolute Percent Error (MAPE): Average

absolute percent error
=
Example: Compute MAD, MSE and MAPE for the following

data, showing actual and predicted numbers of accounts
serviced.
217
215
213
216
216
215
210
214
213
211
219
214
216
217
212
216
Example: Compute MAD, MSE and MAPE for the following

data, showing actual and predicted numbers of accounts
serviced.
217
215
213
216
3
9
1.41%
-3
. 1
1
216 215
1
=
=
= . %0.46%
=
=
=
.
= = = .
210
214
-4
4
16
1.90%
213
211
0.94%
219
214
25
2.28%
216
217
-1
0.46%
212
216
-4
16
1.89%
-2
22
76
10.26%
0.92%
Three commonly used measures for summarizing

historical errors are:
1. MAD: Weights all errors evenly.

=
2. MSE: Weights errors according to their

squared values.
( )
=
3. MAPE: Weights errors according to relative

error.
=
There are two general approaches:
QUALITATIVE APPROACHES: For subjective inputs,

which often oppose precise numerical description. Permit
inclusion of SOFT INFORMATION like human factors,
personal opinions and hunches. These information are
difficult or impossible to quantify.
QUANTITATIVE APPROACHES: Involve either the
projection of historical data or the development of
associative models. They use HARD DATA. Usually avoid
personal opinions.
IN PRACTISE, EITHER OR BOTH
There are
techniques:
three
classes
of
forecasting
JUDGMENTAL FORECASTS
TIME-SERIES FORECAST
ASSOCIATIVE MODLS
JUDGMENTAL FORECASTS
Available
In theseapproaches:
approaches, forecasts
rely solely on
judgment
opinion. For instance:
Executiveand
Opinion
If the forecast is needed quickly.

Salesforce Opinion
If there are some unclear political or economical
conditions
and new data is not available yet.
Consumer Surveys
Introduction of new products or redesign of

Other
Approaches:
Method
existing
products orDelphi
packaging
where there are no
historical data.
A time series is a time-ordered sequence of

observations taken at regular intervals (e.g.,
hourly, daily, weekly, monthly, quarterly,
annually).
Forecasting techniques based on time-series
data are made on the assumption that future
values of the series can be estimated
from past values.
Irregula
r
variatio
n
Common behaviors of time-series:

1. Trend
Trend
2. Seasonality
3. Cycles
Cycles
4. Irregular variations
Seasonal
variations
5. Random variations
90
89
88
TIME-SERIES FORECAST: Naive Methods
A naive forecast uses a single previous

value of a time series as the base of the
forecast.
Examples:
- Stable series: =
- Seasonal variations: =
- Trend: = +( )
TIME-SERIES FORECAST: Averaging Methods

These techniques smooth variations in the data
where there are a great amount of random
variations or White Noise.

1. Moving Average Technique:
+ + +
= =
=
Example: Compute three and five moving averages for the

following data:
Period
Demand
42
40
43
40
41
Answer: = . , = .

2. Weighted Moving Average Technique:
= =
= + + +
=
Example: Compute a weighted average forecast using a

weight of 0.40 for the most recent period, 0.30 for the next
most recent, 0.20 for the next, and 0.10 for the next:
Period
Demand
42
40
43
40
41
Answer: = .

3. Exponential Smoothing Technique:
= 1 + 1 1
= 1 + 1
= 1 1 + 1
Example:
Compute the forecast for the following data using exponential

smoothing technique with = 0.10 and = 0.40.
Actual
= .
= .
Period,
(t)
Demand
Forecast
Forecast
42
40
43
40
41
39
46
44
45
10
38
11
40
12
Example:
Compute the forecast for the following data using exponential

smoothing technique with = 0.10 and = 0.40.
Actual
= .
= .
Period,
(t)
Demand
Forecast
Forecast
42
40
42
42
43
41.8
41.2
40
41.92
41.92
41
41.73
41.15
39
41.66
41.09
46
41.39
40.25
44
41.85
42.55
45
42.07
43.13
10
38
42.35
43.88
11
40
41.92
41.53
41.73
40.92
12
Example:
Compute the error performance of these three forecasting

techniques using the following data:
Naive
Period,
t
Deman
d
42
40
43
40
41
39
46
44
45
10
38
11
40
MAD
MSE
MAPE
Foreca
st
Error
Two-Period MA
Foreca
st
Error
ES
Foreca
st
Error
Example:
Compute the error performance of these three forecasting

techniques using the following data:
Naive
Foreca
st
Two-Period MA
Period,
t
Deman
d
Error
Foreca
st
42
40
42
-2
43
40
41
40
43
-3
41
40
39
Error
ES
Foreca
st
Error
42
-2
41.8
1.2
41.5
-1.5
41.92
-1.92
41.5
-0.5
41.73
-0.73
41
-2
40.5
-1.5
41.66
-2.66
46
39
40
41.39
4.61
44
46
-2
42.5
1.5
41.85
2.15
45
44
45
42.07
2.93
10
38
45
-7
44.5
-6.5
42.36
-4.36
11
40
38
41.5
-1.5
41.92
-1.92
MAD
3.11
2.33
2.50
MSE
16.25
11.44
8.73
MAPE
7.49%
5.64%
5.98%
TIME-SERIES FORECAST: Trend Methods

1. Linear Trend Technique:
= +

=
( )
=
Where n=Number of periods and y=Value of the time series

1. Linear Trend Technique: Example
Using the following data, determine the equation of the trend line and predict for
weeks 11 and 12.
800
Unit Sales (y)
700
724
720
728
740
742
758
750
770
10
775
780
760
Sales
Week (t)
740
720
700
680
0
6
Week
10
12

1. Linear Trend Technique: Example
Using the following data, determine the equation of the trend line and predict for
weeks 11 and 12.
800
Week
Unit Sales
ty
10 41358 55(7407)
(t)
(y)
=
= 7.51
780
10
385
55(55)
1
700
700
724
1448
720
2160
728
2912
740
3700
742
4452
758
5306
750
6000
770
6930
10
775
7750
7407
41358
760
7407 7.51(55)
=
= 699.40
10
Sales
740
720
= 699.40 + 7.51
700
680
0
2 = .4 ,
6
Week
=789.52
8
10
12
Associative Forecasting Techniques:
Identification of related variables that can be

used to predict values of the variable of
interest.
The essence of associative techniques is the
development of an equation that summarizes
the effect of predictor variables.
The primary method of analysis is known as
Regression.
Linear Regression:
Obtain a equation of
a straight line that
Minimizes the sum of
squared vertical
Deviations of data
points from the line
(Least Squares Error)
Linear Regression:
= +
=
( )( )
( ) ( )
or
Linear Regression:
Example: Based on the

following data about a company unit sales and profits,
obtain a regression line and predict profit when sale is $10
million.
= . , = .
= = . + . = .
Linear Regression:
How accurate a prediction might be for a
linear regression line:
Standard Error of Estimates:
=
( )

Linear Regression:
Indicator: Uncontrollable variables that tend to lead or
precede changes in a variable of interest.
Condition for a valid indicator:
1. There should be a logical explanation for the
relations.
2. Movement of the indicator must precede movement
of the dependent variable.
3. A fairly high correlation should exist between the
two variables.

Linear Regression:
Correlation: Measures the strength and direction of
relationship between two variables.
=
( )( )
( 2 ) ( )2
1, +1
2 ( )2
+1:
Direct Relationship
0: No linear relationship
1:
Inverse Relationship
2 : Percentage of the changes in the dependent variable that can be

explained by the independent variable.
Associative Forecasting Techniques: Linear

Regression
60
Unit Sold, Y
Example: Sales of new houses and three-month lagged

50
unemployment
are shown in the following
table.
Determine
= 71.85
6.91
if unemployment
levels can be used =
to 0.966
predict demand for
40
new houses and if so, derive a predictive
and
2 = . equation
%
30
explain the relationship.
20
10
0
3
Level of Unemployment (%), x
10

Demand Forecasting

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Demand Forecasting

Uploaded by

Copyright:

Available Formats

Dr.

1. They assume that the same system that

The forecast should be:

1. Determine the purpose of the forecast.

Accuracy of a forecast is minimizing the

Forecast ERROR is the difference between the

Three commonly used measures for summarizing

2. Mean Squared Error (MSE): Average of squared errors

3. Mean absolute Percent Error (MAPE): Average

Example: Compute MAD, MSE and MAPE for the following

Example: Compute MAD, MSE and MAPE for the following

Three commonly used measures for summarizing

1. MAD: Weights all errors evenly.

2. MSE: Weights errors according to their

3. MAPE: Weights errors according to relative

There are two general approaches:

QUALITATIVE APPROACHES: For subjective inputs,

IN PRACTISE, EITHER OR BOTH

If the forecast is needed quickly.

Introduction of new products or redesign of

A time series is a time-ordered sequence of

Common behaviors of time-series:

TIME-SERIES FORECAST: Naive Methods

A naive forecast uses a single previous

TIME-SERIES FORECAST: Averaging Methods

TIME-SERIES FORECAST: Averaging Methods

Example: Compute three and five moving averages for the

TIME-SERIES FORECAST: Averaging Methods

Example: Compute a weighted average forecast using a

TIME-SERIES FORECAST: Averaging Methods

Compute the forecast for the following data using exponential

Compute the forecast for the following data using exponential

Compute the error performance of these three forecasting

Compute the error performance of these three forecasting

TIME-SERIES FORECAST: Trend Methods

Where n=Number of periods and y=Value of the time series

TIME-SERIES FORECAST: Trend Methods

Unit Sales (y)

TIME-SERIES FORECAST: Trend Methods

Associative Forecasting Techniques:

Identification of related variables that can be

Associative Forecasting Techniques:

Associative Forecasting Techniques:

Associative Forecasting Techniques:

Example: Based on the

Associative Forecasting Techniques:

Associative Forecasting Techniques:

Associative Forecasting Techniques:

2 : Percentage of the changes in the dependent variable that can be

Associative Forecasting Techniques: Linear

Example: Sales of new houses and three-month lagged

Level of Unemployment (%), x

You might also like