c





 

      


á

 Need for forecasting
Basis for most planning decisions
˜ Scheduling
˜ Inventory
˜ Production
˜ Facility layout
˜ Work force
˜ Distribution
˜ Purchasing
˜ Sales
Sources of data for forecasting

˜ Company Records
˜ Published records
˜ Journals
˜ Market Surveys
˜ News papers
Forecasting Model
 ypes of Forecasting Models
V ypes of Forecasts
˜ ²ualitative --- based on experience, judgement, knowledge;
˜ ²uantitative --- based on data, statistics;

V Methods of Forecasting
V time series models (e.g. exponential smoothing) ± trend,
seasonal, cyclical patterns
V causal models (e.g. regression) ± based on relationship
between variable to be forecasted and an independent
variable
 Forecasting Horizons

Production Resource Forecasts

Long range Medium range Short range
Years Months Weeks
Factory Capacities Department capacities, Demand forecasting,
Capital funds Purchased Material staffing levels (labor),
Plant location Aggregate planning, inventory levels,
Product Planning capacity planning,
sales forecasts,

²ualitative ²uantitative, ²uantitative

 Forecasting echniques







   

"
  "c


"
á
 "
 
á

°

 

Historical ime Series
rend Pattern
Cyclical Pattern
Seasonal Pattern
 ypes of Forecasting Models
V ypes of Forecasts
˜ ²ualitative --- based on experience, judgement, knowledge;
˜ ²uantitative --- based on data, statistics;

V Methods of Forecasting
V time series models (e.g. exponential smoothing);
V causal models (e.g. regression) ± based on relationship
between variable to be forecasted and an independent
variable
V Assumptions of ime Series Models
˜ here is information about the past;
˜ his information can be quantified in the form of data;
˜ he pattern of the past will continue into the future.
 Simple Moving Average
V Forecast  is average of  previous observations or
actuals c:
~
› ~ 
›  ›
~  R  › ~

›
~
› ~  


 › ~

V Note that the  past observations are equally weighted.

V Issues with moving average forecasts:
˜ All  past observations treated equally;
˜ 9bservations older than  are not included at all;
˜ Requires that  past observations be retained;
˜ Problem when 1000's of items are being forecast.
 Simple Moving Average
V Include  most recent observations
V Weight equally
V Ignore older observations
V Applied to forecast for only one period into the future






   
 
 Moving Average ± Example 1
c










 






Moving Average ± Example 1
- Solution
Moving Average ± Example 1
- Solution
Moving Average ± Example 1
- Solution
Moving Average ± Example 1
- Solution
 Moving Average ± Example 1
- Solution

Moving Average Forecasting for the 11th

months is 96
 Weighted Moving Average
V Include  most recent observations
V More weight is assigned to the recent demand values
V Ignore older observations
V Applied to forecast for only one period into the future


 c









 





 




   
 
Weighted Moving Average ± Example 2
c










 






Weighted Moving Average
Example 2 - Solution
Weighted Moving Average
Example 2 - Solution
Weighted Moving Average
Example 2 - Solution
Weightd Moving Average
Example 2 - Solution
 Weighted Moving Average
Example 2 - Solution

Moving Average Forecasting for the 9th

months is 81.5
 Exponential Smoothing I
V Include all past observations
V Weight recent observations much more heavily than
very old observations
V Most popular
  







 



  








 


c
  c







 


  á 

 



 Exponential Smoothing I
V Include all past observations
V Weight recent observations much more heavily
than very old observations:



c



 




 

 
 Exponential Smoothing I
V Include all past observations
V Weight recent observations much more heavily
than very old observations:

•Y Y


c



 




 

 
 Exponential Smoothing I
V Include all past observations
V Weight recent observations much more heavily
than very old observations:

•Y Y


c



 




 
%

 
 Exponential Smoothing I
V Include all past observations
V Weight recent observations much more heavily
than very old observations:

•Y Y


c



 




 
%


%

 
Exponential Smoothing: Math
Exponential Smoothing: Math
Exponential Smoothing: Math
Exponential Smoothing: Math



 
c  

 

 

 Exponential Smoothing: Math

›

›
~ Ú

› Ú~
~ Ú

› Ú R
›

› ~ Ú

› Ú~

V hus, new forecast is weighted sum of old forecast and actual

demand
V Notes:
˜ 9nly 2 values (c and  ) are required, compared with  for
moving average
˜ Rule of thumb:  < 0.5
˜ ypically,  = 0.2 or  = 0.3 work well
 Exponential Smoothing ± Example 3




 
 
 




 


























 









 











 




Month Actual 9ld New Forecast

Demand Forecast Forecast Error
Jan 450
Feb 460
Mar 465
Apr 434
May 420
Jun 498
Jul 462
Exponential Smoothing ± Example 3
Month Actual 9ld New Forecast
Demand Forecast Forecast Error
Jan 450 400
Feb 460
Mar 465
Apr 434
May 420
Jun 498
Jul 462



 
c  

 

 


 
Exponential Smoothing ± Example 3
Month Actual 9ld New Forecast
Demand Forecast Forecast Error
Jan 450 400
Feb 460 410
Mar 465
Apr 434
May 420
Jun 498
Jul 462



 
c 


 

 



  



  




Exponential Smoothing ± Example 3
Month Actual 9ld New Forecast
Demand Forecast Forecast Error
Jan 450 400
Feb 460 410 410
Mar 465
Apr 434
May 420
Jun 498
Jul 462



 
c 


 

 




  



  




Exponential Smoothing ± Example 3
Month Actual 9ld New Forecast
Demand Forecast Forecast Error
Jan 450 400
Feb 460 410 410
Mar 465 420
Apr 434
May 420
Jun 498
Jul 462



 
c 


 

 




  



  




Exponential Smoothing ± Example 3
Month Actual 9ld New Forecast
Demand Forecast Forecast Error
Jan 450 400 - + 50
Feb 460 410 410 + 50
Mar 465 420 420 + 45
Apr 434 429 429 +5
May 420 430 430 a10

Jun 498 428 428 + 70

Jul 462 442 442 + 20



 
c 


 

 




  



  




 ¯
 
Double Exponential Smoothing
or
Exponential Smoothing with rend
or
Adjusted Exponential Smoothing
 ¯
 
Double Exponential Smoothing
V Ideas behind smoothing with trend:
˜ ½½De-trend'' time-series by separating =
from   effects
˜ Smooth =
in usual manner using À
˜ Smooth   forecasts in usual manner using p
V Smooth the =
forecast 
O
  
c
 

 

O




V Smooth the  forecast 
   
O
  O
  
 
 

V Forecast periods into future  with =
and  


O
 


Exponential Smoothing with rend
± Example 4
°  



 






 














  
























 








! "
   
 

Week Demand Week Demand

Jan 1 650 Feb 1 625

2 600 2 675

3 550 3 700

4 650 4 710
 Exponential Smoothing with rend
± Example 4
! "
   
  Week Demand Week Demand
Jan 1 650 Feb 1 625


2 600 2 675

 3 550 3 700
4 650 4 710

O
  
c
 

 

O





   
O
  O
  
 
 




O
 



 Exponential Smoothing with rend
± Example 4
! "
   
  B t-1 D t-1 Bt t F t+1

We Pre Act Smooth Smooth Next



ek Avg Demand Avg rend Project
ion


Jan 1 600 650 605
2 600
O
  
c
 

 

O




3 550

  


  




 4 650
   
O
  O
  
 
 
 Feb 1 625
2 675
3 700


O
 


4 710
 Exponential Smoothing with rend
± Example 4
! "
   
  B t-1 D t-1 Bt t F t+1

We Pre Act Smooth Smooth Next



ek Avg Demand Avg rend Project
ion


Jan 1 600 650 605 1.0
2 600
O
  
c
 

 

O




3 550

  


  




 4 650
   
O
  O
  
 
 
 Feb 1 625

  

 

 
 

  2 675

3 700


O
 


4 710
 Exponential Smoothing with rend
± Example 4
! "
   
  B t-1 D t-1 Bt t F t+1

We Pre Act Smooth Smooth Next



ek Avg Demand Avg rend Project
ion


Jan 1 600 650 605 1.0 606
2 605 600
O
  
c
 

 

O




3 550

  


  




  4 650
   
O
  O
  
 
 
 Feb 1 625

  
 
 

 
 

  2 675

3 700


O
 


4 710

 


 Exponential Smoothing with rend
± Example 4
! "
   
  B t-1 D t-1 Bt t F t+1

We Pre Act Smooth Smooth Next



ek Avg Demand Avg rend Project
ion


Jan 1 600 650 605 1.0 606
2 605 600 605.4
O
  
c
 

 

O




3 550

  


  




  4 650
   
O
  O
  
 
 
 Feb 1 625
2 675
3 700


O
 


4 710
 Exponential Smoothing with rend
± Example 4
! "
   
  B t-1 D t-1 Bt t F t+1

We Pre Act Smooth Smooth Next



ek Avg Demand Avg rend Project
ion


Jan 1 600 650 605 1.0 606
2 605 600 605.4 0.88
O
  
c
 

 

O




3 550

  


  




  4 650
   
O
  O
  
 
 
 Feb 1 625

  
 
 

 
 

  2 675

3 700


O
 


4 710
 Exponential Smoothing with rend
± Example 4
! "
   
  W B t-1 D t-1 Bt t F t+1
e
Pre Act Smooth Smooth Next

 e
Avg Demand Avg rend Projection
k


Jan 1 600 650 605 1.0 606
2 605 600 605.4 0.88 606.38
O
  
c
 

 

O



 3 550

  


  




  4 650

   
O
  O
  
 
 


Feb 1 625
2 675

  
 
 

 
 

 
3 700


O
 


4 710

  
 
 
 Exponential Smoothing with rend
± Example 4
! "
   
  W B t-1 D t-1 Bt t F t+1
e
Pre Act Smooth Smooth Next

 e
Avg Dema Avg rend Projec
k nd

 tion
Jan 1 600 650 605 1.0 606

2 605 600 605.4 0.88 606.38

O
  
c
 

 

O




3 605.4 550 600.65 -0.246 600.40
4 600.65 650 605.36 0.742 606.10
   
O
  O
  
 
 
 Feb 1 605.36 625 607.99 1.120 609.11

2 607.99 675 615.70 2.44 618.14

3 615.70 700 626.33 4.08 630.4



O
 


4 626.33 710 738.37 5.67 644.04
 Forecasting Performance
¯
  
 
V Mean Forecast Error (MFE or Bias): Measures
average deviation of forecast from actuals.

V Mean Absolute Deviation (MAD): Measures

average absolute deviation of forecast from
actuals.

V Mean Absolute Percentage Error (MAPE):

Measures absolute error as a percentage of the
forecast.

V Standard Squared Error (MSE): Measures

variance of forecast error
 Mean Forecast Error (MFE or Bias)




›

› Ú
› <

V Want MFE to be as close to zero as possible --

minimum bias
V A large positive (negative) MFE means that the
forecast is undershooting (overshooting) the actual
observations
V Note that zero MFE does not imply that forecasts are
perfect (no error) -- only that mean is ³on target´
V Also called forecast BIAS
 Mean Absolute Deviation (MAD)

~ 
 Ã
 ~


V Measures absolute error

V Positive and negative errors thus do not cancel out (as
with MFE)
V Want MAD to be as small as possible
V No way to know if MAD error is large or small in
relation to the actual data
 Mean Absolute Percentage Error
(MAPE)


~




Ã
~

V Same as MAD, except ...

V Measures deviation as a percentage of actual data
 Mean Squared Error (MSE)



›

› Ú
› < d

V Measures squared forecast error -- error variance

V Recognizes that large errors are disproportionately more
³expensive´ than small errors
V But is not as easily interpreted as MAD, MAPE -- not as
intuitive
Simple Regression / Linear
Regression
 á
  
   

 

 

  
  
 

!   !  
"#   á 

˜

c 

   
   
 
  

 


    



á
  
   

=
 = 

 = 

 

p

    

  

 

 


 

  


  






  

á ×$
  

i

 



  
   

  




  


 
 


   
 

   
o find the value of a, b use the formulae below

™Y = n * a + b ™X

™XY = n ™X + b ™X2


# 


 #

#

a = -------------------------
n * 
#  #

#

 #


b = --------------------



#  #

Find #





#



#2

%&'(& á

Reed Auto periodically has a special week-long sale. As part of the

advertising campaign Reed runs one or more television
commercials during the weekend preceding the sale. Data from a
sample of 5 previous sales showing the number of  ads run and
the number of cars sold in each sale are shown below.

!
 
 !
 
1 14
3 24
2 18
1 17
3 27

No. of  ads ± Independent ariable ± X

No of cars sold ± Dependent ariable - Y

%&'(& á







 
 
 


%&'(& á
)
~ ~


~

~ ~

 

) ) )*
~ ~ ~ ~

 

~

~ ~ ~ ~

  ~ 
 
%&'(& á

~ ~ ~ ~

 

~

~ ~ ~ ~

  ~ 
 +,, +, , -

%&'(& á

V Slope for the Estimated Regression Equation

= = ?
-Intercept for the Estimated Regression Equation
 = ?
V Estimated Regression Equation
Y=?
Fortunately, there is software...