Infodays-04 Fcst Models - Archive...historical periods Weighted Moving Average (14) Weighted Moving...

Forecasting Workshop

11/17/2003

SAP AG 2002, Title of Presentation, Speaker Name / 2

Background

What it is: A detailed presentation focusing on the Mathematical methods of prediction that are standard in APO-Demand Planning.

It is based on the Functionality found in release 3.1.

Anyone interested in upgrading it to 4.0 ?


Overview of Forecasting in APO

333

111

Univariate Forecast Basics

222 History

444 Univariate Models in APOModels in APOModels in APOModels in APO

555 Forecast Error Analysis

666

777

Master Forecast Profile Setup

Forecast Execution

Overview of the Forecasting Workshop


333 Univariate Forecast Basics



What we will talk about today


Elements of the Univariate Forecast

�Actual Value

�Basic Value

�Trend Value

�Seasonal Index

�Length of Season

�Initialization

�Ex-post Forecast

�Forecast

�Parameters

Exceptions are models 13, 14, 35, 70, 80, 94


Actual Value

Data

Valu

es

Past Horizon

Historical values used as the basis of theforecast


Basic Value

Data

Valu

es

Forecast Horizon

Basic Value - Controls Vertical Placement

1

22000

1000


Trend Value

Data

Valu

es

Forecast Horizon

Trend - Slope of line

1

2

.75

-.75


Seasonal Index

Data

Valu

es

Forecast Horizon

Seasonal index - Divergence from Basic Value

1.00

2.50

.25


Length of Season

Data

Valu

es

Number of Periods in the Season

Typically = one year

Season


Initialization

Basic value, Trend value, Seasonal indices, and the Mean Absolute Deviation are initialized.


Ex-post Forecast

An Ex post Forecast is a forecast run in the past for comparisonAn Ex post Forecast is a forecast run in the past for comparisonAn Ex post Forecast is a forecast run in the past for comparisonAn Ex post Forecast is a forecast run in the past for comparisonagainst historical values. The exagainst historical values. The exagainst historical values. The exagainst historical values. The ex----post forecast is used to measure post forecast is used to measure post forecast is used to measure post forecast is used to measure model fit.model fit.model fit.model fit.

Data

Valu

es

Future

Forecast

History

Initialization Ex-post Forecast

History Values

ForecastValues

- Initialization length is dependent on the forecast model- Ex-post forecast length is the history horizon less the

initialization horizon


Ex-post Forecast

The following Models do not generate an ExThe following Models do not generate an ExThe following Models do not generate an ExThe following Models do not generate an Ex----post forecast:post forecast:post forecast:post forecast:

Manual forecastManual forecastManual forecastManual forecast

Historical data adoptedHistorical data adoptedHistorical data adoptedHistorical data adopted

Weighted moving averageWeighted moving averageWeighted moving averageWeighted moving average

Moving averageMoving averageMoving averageMoving average

70707070

60606060

14141414

13131313


General Calculation

Basic(t), trend(t), seasonal(t)

=

Function(actual(t), basic(t-1), trend(t-1), seasonal(t-1s))

Forecast = Function(basic(t), trend(t), seasonal(t))


Trend Model Example – Initialization & Ex-Post

Period 1 2 3 4 5 6 7 1 2 3 4

History 80 95 100 115 125 110 110

G - Basic Value 102

T - Trend 10.0

EX Ex-post

F Forecast

History Horizon Forecast Horizon

Basic Value Initialization

Trend Initialization

Ex-Post Forecast

Ex(t +i) = G(t) + i * T(t)

Where i = 1


Period 1 2 3 4 5 6 7 1 2 3 4

History 80 95 100 115 125 110 110

G - Basic Value 102 113

T - Trend 10.0 10.3

EX Ex-post 112 123

F Forecast


Trend Model Example – Calculate New Basic and Trend

Trend Value T(t) = ββββ [ G(t) - G(t-1) ] + (1-ββββ)T(t-1)

10.3 = .3 [ 113 – 102 ] + (.7) 10.0

Basic Value G(t) = ααααV(t) + (1- αααα) [ G(t-1) + T(t-1) ]

113 = .3 (115) + (.7) [ 102 + 10.0 ]

Ex-Post Forecast Ex(t +i) = G(t) + i * T(t)Where i = 1

123 = 113 + 10.3


Period 1 2 3 4 5 6 7 1 2 3 4

History 80 95 100 115 125 110 110

G - Basic Value 102 113

T - Trend 10.0 10.3

EX Ex-post 112 123

F Forecast


Trend Model Example – Calculation Repeated


124 127

10.5 8.3

123 134 135

124

10.5

123 134

124 127 128

10.5 8.3 6.1

123 134 135

134

Forecast P(t+i) = G(t) + i * T(t)




Period 1 2 3 4 5 6 7 1 2 3 4

History 80 95 100 115 125 110 110

G - Basic Value 102 113 124 127 128 128

T - Trend 10.0 10.3 10.5 8.3 6.1 6.1

EX Ex-post 112 123 134 135

F Forecast 134 140


Trend Model Example – Calculation Repeated


Forecast P(t+i) = G(t) + i * T(t)

110

128 128 128

6.1 6.1 6.1

135

134 140 146

1

110

128 128 128 128

6.1 6.1 6.1 6.1

135

134 140 146 152

2




Model parameters

The model parameters, Alpha, Beta, and Gamma, control the weighting of each historical values.

�Alpha is used in the basic value calculation

�Beta is used in trend value calculation

�Gamma is used in the Seasonal index calculation

The value for the parameters range from 0 to 1. A higher value will place more emphasis on recent history.

The parameters also control how reactive the forecast is to changes in historical patterns.


Univariate Forecasting Models

� Constant ModelsConstant ModelsConstant ModelsConstant Models

� Moving Average Models Moving Average Models Moving Average Models Moving Average Models

� Trend ModelsTrend ModelsTrend ModelsTrend Models

� Seasonal ModelsSeasonal ModelsSeasonal ModelsSeasonal Models

� Seasonal Trend ModelsSeasonal Trend ModelsSeasonal Trend ModelsSeasonal Trend Models

� Automatic Model Selection I and IIAutomatic Model Selection I and IIAutomatic Model Selection I and IIAutomatic Model Selection I and II

� Sporadic DemandSporadic DemandSporadic DemandSporadic Demand

� Historical Data ModelHistorical Data ModelHistorical Data ModelHistorical Data Model

� Linear RegressionLinear RegressionLinear RegressionLinear Regression

� User ExitUser ExitUser ExitUser Exit


Constant Models

3 models within the constant model group:3 models within the constant model group:3 models within the constant model group:3 models within the constant model group:

Constant Model (10)Constant Model (10)Constant Model (10)Constant Model (10)

1111stststst Order Exponential Smoothing (11)Order Exponential Smoothing (11)Order Exponential Smoothing (11)Order Exponential Smoothing (11)

Constant Model with automatic Alpha adaptation (12)Constant Model with automatic Alpha adaptation (12)Constant Model with automatic Alpha adaptation (12)Constant Model with automatic Alpha adaptation (12)

Constant Models follow the pattern:Constant Models follow the pattern:Constant Models follow the pattern:Constant Models follow the pattern:

Same Formula !10 = 11


Constant Model – 1st order Exponential Smoothing (10,11)

Use this model when there are no seasonal variations or trend Use this model when there are no seasonal variations or trend Use this model when there are no seasonal variations or trend Use this model when there are no seasonal variations or trend patterns. Alpha parameter controls the weighting of history in patterns. Alpha parameter controls the weighting of history in patterns. Alpha parameter controls the weighting of history in patterns. Alpha parameter controls the weighting of history in determining the basic value.determining the basic value.determining the basic value.determining the basic value.

� Basic value and exBasic value and exBasic value and exBasic value and ex----post forecast are equalpost forecast are equalpost forecast are equalpost forecast are equal� Higher alpha puts more weight on recent valuesHigher alpha puts more weight on recent valuesHigher alpha puts more weight on recent valuesHigher alpha puts more weight on recent values� Higher alpha makes the forecast more reactive to recent changes Higher alpha makes the forecast more reactive to recent changes Higher alpha makes the forecast more reactive to recent changes Higher alpha makes the forecast more reactive to recent changes in in in in

historyhistoryhistoryhistory


Example of a Constant Model


Constant Model – Automatic Adaptation of the Alpha factor (12)

The system automatically adjusts the alpha factor based on the iThe system automatically adjusts the alpha factor based on the iThe system automatically adjusts the alpha factor based on the iThe system automatically adjusts the alpha factor based on the initial nitial nitial nitial alpha and the smallest tracking signal.alpha and the smallest tracking signal.alpha and the smallest tracking signal.alpha and the smallest tracking signal.


Moving Average Models

Two models are within the Moving Average groupTwo models are within the Moving Average groupTwo models are within the Moving Average groupTwo models are within the Moving Average group

Moving Average (13)Moving Average (13)Moving Average (13)Moving Average (13) -------- the forecast equals the average of the the forecast equals the average of the the forecast equals the average of the the forecast equals the average of the historical periodshistorical periodshistorical periodshistorical periods

Weighted Moving Average (14) Weighted Moving Average (14) Weighted Moving Average (14) Weighted Moving Average (14) –––– the forecast equals the weighted the forecast equals the weighted the forecast equals the weighted the forecast equals the weighted average of the historical periods. The weighting is defined viaaverage of the historical periods. The weighting is defined viaaverage of the historical periods. The weighting is defined viaaverage of the historical periods. The weighting is defined via a a a a weighting profile.weighting profile.weighting profile.weighting profile.

Period Weighting

-5 10%

-4 20%

-3 30%

-2 40%

-1 50%


Moving Average-Formulas

Moving Average-13 Weighted Moving Average-14

No ex-post forecast calculated


Trend Models

There are four models for data with trend:There are four models for data with trend:There are four models for data with trend:There are four models for data with trend:

Forecast with Trend Model (20)Forecast with Trend Model (20)Forecast with Trend Model (20)Forecast with Trend Model (20)

HoltHoltHoltHolt’’’’s Method (21) s Method (21) s Method (21) s Method (21)

Second Order Exponential Smoothing (22) Second Order Exponential Smoothing (22) Second Order Exponential Smoothing (22) Second Order Exponential Smoothing (22)

Trend model with Automatic Alpha Adaptation (2nd order) (23) Trend model with Automatic Alpha Adaptation (2nd order) (23) Trend model with Automatic Alpha Adaptation (2nd order) (23) Trend model with Automatic Alpha Adaptation (2nd order) (23)



Holt’s Method (20, 21)

Basic Value Basic Value Basic Value Basic Value G(t) = ααααV(t) + (1-αααα) [ G(t-1) + T(t-1) ]

Trend ValueTrend ValueTrend ValueTrend Value T(t) = ββββ [(G(t) - G(t-1) ] + (1-ββββ)T(t-1)

Forecast value for period (t +1) Forecast value for period (t +1) Forecast value for period (t +1) Forecast value for period (t +1) P(t+i) = G(t) + i * T(t)

Formulas


Second Order Exponential Smoothing (22)

Encompasses a double smoothing method where the basic value of Encompasses a double smoothing method where the basic value of Encompasses a double smoothing method where the basic value of Encompasses a double smoothing method where the basic value of the first calculation is used as the input for the second. Thethe first calculation is used as the input for the second. Thethe first calculation is used as the input for the second. Thethe first calculation is used as the input for the second. The result is a result is a result is a result is a basic value that is more reactive to changes in demand while basic value that is more reactive to changes in demand while basic value that is more reactive to changes in demand while basic value that is more reactive to changes in demand while minimizing the trend effect. minimizing the trend effect. minimizing the trend effect. minimizing the trend effect.


Trend with Automatic Alpha Adaptation (2nd order) (23)

This is the same as Second Order Exponential Smoothing except thThis is the same as Second Order Exponential Smoothing except thThis is the same as Second Order Exponential Smoothing except thThis is the same as Second Order Exponential Smoothing except the e e e systems determines the Alpha value. The minimum Alpha is 0.05 ansystems determines the Alpha value. The minimum Alpha is 0.05 ansystems determines the Alpha value. The minimum Alpha is 0.05 ansystems determines the Alpha value. The minimum Alpha is 0.05 and d d d the maximum is 0.90. The system determines the Alpha factor thrthe maximum is 0.90. The system determines the Alpha factor thrthe maximum is 0.90. The system determines the Alpha factor thrthe maximum is 0.90. The system determines the Alpha factor through ough ough ough an iterative processes of comparing the MAD to the Error Total (an iterative processes of comparing the MAD to the Error Total (an iterative processes of comparing the MAD to the Error Total (an iterative processes of comparing the MAD to the Error Total (ET). ET). ET). ET). (Same Method as Constant with Automatic Alpha Adaptation).(Same Method as Constant with Automatic Alpha Adaptation).(Same Method as Constant with Automatic Alpha Adaptation).(Same Method as Constant with Automatic Alpha Adaptation).


Trend Dampening Profile

As the historical values do not show a decrease in trend as the As the historical values do not show a decrease in trend as the As the historical values do not show a decrease in trend as the As the historical values do not show a decrease in trend as the market market market market matures, a Trend Dampening profile can be used to smooth out thematures, a Trend Dampening profile can be used to smooth out thematures, a Trend Dampening profile can be used to smooth out thematures, a Trend Dampening profile can be used to smooth out thetrend. trend. trend. trend.


Seasonal Models

Seasonal Models are used to depict a Seasonal forecast or a recuSeasonal Models are used to depict a Seasonal forecast or a recuSeasonal Models are used to depict a Seasonal forecast or a recuSeasonal Models are used to depict a Seasonal forecast or a recurring rring rring rring pattern, without any upwards or downwards slope. The pattern ocpattern, without any upwards or downwards slope. The pattern ocpattern, without any upwards or downwards slope. The pattern ocpattern, without any upwards or downwards slope. The pattern occurs curs curs curs across its season or a number of Seasonal Periods (Usually across its season or a number of Seasonal Periods (Usually across its season or a number of Seasonal Periods (Usually across its season or a number of Seasonal Periods (Usually recommended to be 12 Months). The trend value and Beta in this recommended to be 12 Months). The trend value and Beta in this recommended to be 12 Months). The trend value and Beta in this recommended to be 12 Months). The trend value and Beta in this case are zero, signifying no upward or downward slope. case are zero, signifying no upward or downward slope. case are zero, signifying no upward or downward slope. case are zero, signifying no upward or downward slope.


Seasonal Models

There are three Models associated with the seasonal patterns.There are three Models associated with the seasonal patterns.There are three Models associated with the seasonal patterns.There are three Models associated with the seasonal patterns.

Forecast with the Seasonal Model (30) Forecast with the Seasonal Model (30) Forecast with the Seasonal Model (30) Forecast with the Seasonal Model (30)

Seasonal Trend based on the WinterSeasonal Trend based on the WinterSeasonal Trend based on the WinterSeasonal Trend based on the Winter’’’’s Method (31) s Method (31) s Method (31) s Method (31)

Season + Linear Regression (35)Season + Linear Regression (35)Season + Linear Regression (35)Season + Linear Regression (35)



Forecast with the Seasonal Model (30)

This model uses first order exponential smoothing where both AlpThis model uses first order exponential smoothing where both AlpThis model uses first order exponential smoothing where both AlpThis model uses first order exponential smoothing where both Alpha ha ha ha and Gamma are assigned.and Gamma are assigned.and Gamma are assigned.and Gamma are assigned.


Season + Linear Regression (35)

�Seasonal indices calculated

�Seasonality removed from data

�Forecast created using linear regression

�Seasonality added back to forecast


Seasonal Trend Models

There are two Models for the Seasonal Trend method:There are two Models for the Seasonal Trend method:There are two Models for the Seasonal Trend method:There are two Models for the Seasonal Trend method:

Forecast with Seasonal Trend Model (40)Forecast with Seasonal Trend Model (40)Forecast with Seasonal Trend Model (40)Forecast with Seasonal Trend Model (40)

First Order Exponential Smoothing (41) First Order Exponential Smoothing (41) First Order Exponential Smoothing (41) First Order Exponential Smoothing (41) Same Formula !

40 = 41

Also know as: Holt-Winters Method

And Triple Exponential Smoothing


First Order Exponential Smoothing (40, 41)

This Model encompasses the same Formula as 20,21,30,31 and uses This Model encompasses the same Formula as 20,21,30,31 and uses This Model encompasses the same Formula as 20,21,30,31 and uses This Model encompasses the same Formula as 20,21,30,31 and uses assigned values for the Alpha, Beta and Gamma Coefficients.assigned values for the Alpha, Beta and Gamma Coefficients.assigned values for the Alpha, Beta and Gamma Coefficients.assigned values for the Alpha, Beta and Gamma Coefficients.


Auto Selection Models

Auto Selection 1 (50)Auto Selection 1 (50)Auto Selection 1 (50)Auto Selection 1 (50)

Test for Trend and Season (53)Test for Trend and Season (53)Test for Trend and Season (53)Test for Trend and Season (53)

Test for Trend (51)Test for Trend (51)Test for Trend (51)Test for Trend (51)

Test for Season (52)Test for Season (52)Test for Season (52)Test for Season (52)

Seasonal model plus test for Trend (54)Seasonal model plus test for Trend (54)Seasonal model plus test for Trend (54)Seasonal model plus test for Trend (54)

Trend Model plus test for Seasonal Pattern (55)Trend Model plus test for Seasonal Pattern (55)Trend Model plus test for Seasonal Pattern (55)Trend Model plus test for Seasonal Pattern (55)

Auto Model Selection procedure 2 (56)Auto Model Selection procedure 2 (56)Auto Model Selection procedure 2 (56)Auto Model Selection procedure 2 (56)



Auto Selection 1 (50, 53)

Tests for Both Seasonal and Trend Patterns.Tests for Both Seasonal and Trend Patterns.Tests for Both Seasonal and Trend Patterns.Tests for Both Seasonal and Trend Patterns.

Seasonal Test: The Auto Correlation Coefficient is Calculated, tSeasonal Test: The Auto Correlation Coefficient is Calculated, tSeasonal Test: The Auto Correlation Coefficient is Calculated, tSeasonal Test: The Auto Correlation Coefficient is Calculated, the he he he Pattern is determined to be Seasonal if the Auto Correlation Pattern is determined to be Seasonal if the Auto Correlation Pattern is determined to be Seasonal if the Auto Correlation Pattern is determined to be Seasonal if the Auto Correlation Coefficient is Greater than 0.3. Coefficient is Greater than 0.3. Coefficient is Greater than 0.3. Coefficient is Greater than 0.3.


Auto Selection 1 (50, 53)

A Trend Significance test is carried outA Trend Significance test is carried outA Trend Significance test is carried outA Trend Significance test is carried out

If both are determined to be significant, a Seasonal Trend ModelIf both are determined to be significant, a Seasonal Trend ModelIf both are determined to be significant, a Seasonal Trend ModelIf both are determined to be significant, a Seasonal Trend Model is is is is used.used.used.used.

If neither tests is determined to be Significant, then the ConstIf neither tests is determined to be Significant, then the ConstIf neither tests is determined to be Significant, then the ConstIf neither tests is determined to be Significant, then the Constant ant ant ant Model is chosen as a defaultModel is chosen as a defaultModel is chosen as a defaultModel is chosen as a default


Test for Trend (51)

Used to Test only for Trend.Used to Test only for Trend.Used to Test only for Trend.Used to Test only for Trend.

This follows the Trend Significance test previously defined. IfThis follows the Trend Significance test previously defined. IfThis follows the Trend Significance test previously defined. IfThis follows the Trend Significance test previously defined. If no no no no Significance is found then a Constant Model is Chosen.Significance is found then a Constant Model is Chosen.Significance is found then a Constant Model is Chosen.Significance is found then a Constant Model is Chosen.


Test for Season (52)

Used to Test only for Seasonal Pattern. Used to Test only for Seasonal Pattern. Used to Test only for Seasonal Pattern. Used to Test only for Seasonal Pattern.

This follows the This follows the This follows the This follows the AutoCorrelationAutoCorrelationAutoCorrelationAutoCorrelation Coefficient Test and Coefficient Test and Coefficient Test and Coefficient Test and choseschoseschoseschoses the the the the Seasonal Model if the value is above 0.3. If not a Constant ModSeasonal Model if the value is above 0.3. If not a Constant ModSeasonal Model if the value is above 0.3. If not a Constant ModSeasonal Model if the value is above 0.3. If not a Constant Model is el is el is el is chosen.chosen.chosen.chosen.


Seasonal model plus test for Trend (54)

Assumes Seasonal pattern exists and tests for Trend pattern. IfAssumes Seasonal pattern exists and tests for Trend pattern. IfAssumes Seasonal pattern exists and tests for Trend pattern. IfAssumes Seasonal pattern exists and tests for Trend pattern. If a a a a trend significance is found, then a Seasonal Trend Model is usedtrend significance is found, then a Seasonal Trend Model is usedtrend significance is found, then a Seasonal Trend Model is usedtrend significance is found, then a Seasonal Trend Model is used. If . If . If . If no significance is found than a Seasonal Model is used.no significance is found than a Seasonal Model is used.no significance is found than a Seasonal Model is used.no significance is found than a Seasonal Model is used.


Trend Model plus test for Seasonal Pattern (55)

Assumes Trend exists and test for a Seasonal Pattern. If the Assumes Trend exists and test for a Seasonal Pattern. If the Assumes Trend exists and test for a Seasonal Pattern. If the Assumes Trend exists and test for a Seasonal Pattern. If the Autocorrelation Coefficient is greater than 0.3 than a Seasonal Autocorrelation Coefficient is greater than 0.3 than a Seasonal Autocorrelation Coefficient is greater than 0.3 than a Seasonal Autocorrelation Coefficient is greater than 0.3 than a Seasonal Trend Trend Trend Trend Model is used, if the Seasonal test is not found to be SignificaModel is used, if the Seasonal test is not found to be SignificaModel is used, if the Seasonal test is not found to be SignificaModel is used, if the Seasonal test is not found to be Significant than a nt than a nt than a nt than a Trend Model is used.Trend Model is used.Trend Model is used.Trend Model is used.


Summary of the Automodel 1 selection process

This shows the Model chosen under the following Conditions:This shows the Model chosen under the following Conditions:This shows the Model chosen under the following Conditions:This shows the Model chosen under the following Conditions:Automodel If:Trend If: Season Then:Uses Model

40 - Seasonal Trend

20 - Trend

Yes

No

Yes by

Default

55 - Trend Model plus test

for Season

30 - Season

40 - Seasonal Trend

Yes by

Default

No

Yes

54 - Seasonal Model plus

test for Trend

30 - Season

10 - Constant

Yes

No

N/A52 - Test for Season

30 - Trend

10 - Constant

N/AYes

No

51 - Test for Trend

10 - Constant

20 - Trend

30 - Season

40 - Seasonal Trend

No

No

Yes

Yes

No

Yes

No

Yes

50 - Auto Selection 1

or

53 - Test for Trend and

Season


Auto Model Selection procedure 2 (56)

Tests for Constant, Seasonal and/or Trend models while attemptinTests for Constant, Seasonal and/or Trend models while attemptinTests for Constant, Seasonal and/or Trend models while attemptinTests for Constant, Seasonal and/or Trend models while attempting to g to g to g to minimize Error by adjusting Alpha, Beta and Gamma..minimize Error by adjusting Alpha, Beta and Gamma..minimize Error by adjusting Alpha, Beta and Gamma..minimize Error by adjusting Alpha, Beta and Gamma..

This model adjusts Alpha, Beta and Gamma between the values of 0This model adjusts Alpha, Beta and Gamma between the values of 0This model adjusts Alpha, Beta and Gamma between the values of 0This model adjusts Alpha, Beta and Gamma between the values of 0.1 .1 .1 .1 to 0.5 in 0.1 increments while calculating the Mean Absolute Devto 0.5 in 0.1 increments while calculating the Mean Absolute Devto 0.5 in 0.1 increments while calculating the Mean Absolute Devto 0.5 in 0.1 increments while calculating the Mean Absolute Deviation iation iation iation (MAD) for each Iteration. The iteration with the Lowest MAD is (MAD) for each Iteration. The iteration with the Lowest MAD is (MAD) for each Iteration. The iteration with the Lowest MAD is (MAD) for each Iteration. The iteration with the Lowest MAD is the the the the Model chosen with the appropriate smoothing factors. Model chosen with the appropriate smoothing factors. Model chosen with the appropriate smoothing factors. Model chosen with the appropriate smoothing factors.


Comparision between 50 and 56 (Auto model 1 & 2)

Auto Selection Model 1 Auto Selection Model 2Auto Selection Model 1 Auto Selection Model 2Auto Selection Model 1 Auto Selection Model 2Auto Selection Model 1 Auto Selection Model 2

Chooses Best Fit of Constant, Trend, Chooses Best Fit of Constant, Trend, Chooses Best Fit of Constant, Trend, Chooses Best Fit of Constant, Trend, Seasonal, Seasonal Trend.Seasonal, Seasonal Trend.Seasonal, Seasonal Trend.Seasonal, Seasonal Trend.

If no Pattern Detected defaults Constant If no Pattern Detected defaults Constant If no Pattern Detected defaults Constant If no Pattern Detected defaults Constant ModelModelModelModel

Resource Intensive (Not Resource Intensive (Not Resource Intensive (Not Resource Intensive (Not recommended for Production use)recommended for Production use)recommended for Production use)recommended for Production use)

Not as precise as Auto Model 2Not as precise as Auto Model 2Not as precise as Auto Model 2Not as precise as Auto Model 2

Determines Best Fit through through Determines Best Fit through through Determines Best Fit through through Determines Best Fit through through Iterative approach of adjusting Iterative approach of adjusting Iterative approach of adjusting Iterative approach of adjusting smoothing factors (Alpha, Beta, smoothing factors (Alpha, Beta, smoothing factors (Alpha, Beta, smoothing factors (Alpha, Beta, Gamma) and determining Smallest Gamma) and determining Smallest Gamma) and determining Smallest Gamma) and determining Smallest Mean Absolute Deviation (MAD)Mean Absolute Deviation (MAD)Mean Absolute Deviation (MAD)Mean Absolute Deviation (MAD)

Determines Seasonal and Trend Determines Seasonal and Trend Determines Seasonal and Trend Determines Seasonal and Trend Significance through Formulas for Significance through Formulas for Significance through Formulas for Significance through Formulas for Auto Correlation (Seasonal Auto Correlation (Seasonal Auto Correlation (Seasonal Auto Correlation (Seasonal Significance) and Trend Significance) and Trend Significance) and Trend Significance) and Trend Significance TestsSignificance TestsSignificance TestsSignificance Tests


The Croston Model (80)

The Croston Model (80) is used when demand is sporadic or The Croston Model (80) is used when demand is sporadic or The Croston Model (80) is used when demand is sporadic or The Croston Model (80) is used when demand is sporadic or intermittent. The return is a Constant model to meet the possibintermittent. The return is a Constant model to meet the possibintermittent. The return is a Constant model to meet the possibintermittent. The return is a Constant model to meet the possible le le le sporadic demand.sporadic demand.sporadic demand.sporadic demand.


Historical Data Model (60)

The Historical Data Model (60) allows you to use your HistoricalThe Historical Data Model (60) allows you to use your HistoricalThe Historical Data Model (60) allows you to use your HistoricalThe Historical Data Model (60) allows you to use your Historical values values values values as your forecast. as your forecast. as your forecast. as your forecast.


Manual Forecast (70)

The Manual Forecast (70) allows the planner to set their own valThe Manual Forecast (70) allows the planner to set their own valThe Manual Forecast (70) allows the planner to set their own valThe Manual Forecast (70) allows the planner to set their own values to ues to ues to ues to generate a forecast. The values chosen are all the values that generate a forecast. The values chosen are all the values that generate a forecast. The values chosen are all the values that generate a forecast. The values chosen are all the values that make make make make up the forecast these are the Smoothing coefficients Alpha, Betup the forecast these are the Smoothing coefficients Alpha, Betup the forecast these are the Smoothing coefficients Alpha, Betup the forecast these are the Smoothing coefficients Alpha, Beta, a, a, a, Gamma (from the Profile) and the Basic, Trend value and SeasonGamma (from the Profile) and the Basic, Trend value and SeasonGamma (from the Profile) and the Basic, Trend value and SeasonGamma (from the Profile) and the Basic, Trend value and Seasonal al al al Index. This model should not be used for Background Processing.Index. This model should not be used for Background Processing.Index. This model should not be used for Background Processing.Index. This model should not be used for Background Processing.


No Forecast (98)

No forecast is conductedNo forecast is conductedNo forecast is conductedNo forecast is conducted


External Forecast (99)

It is also possible to define your own Forecasting model if the It is also possible to define your own Forecasting model if the It is also possible to define your own Forecasting model if the It is also possible to define your own Forecasting model if the model model model model you commonly use is not provided. The external Forecasting featyou commonly use is not provided. The external Forecasting featyou commonly use is not provided. The external Forecasting featyou commonly use is not provided. The external Forecasting feature ure ure ure allows you to use ABAP to code your own forecasting modelallows you to use ABAP to code your own forecasting modelallows you to use ABAP to code your own forecasting modelallows you to use ABAP to code your own forecasting model.


Trend, Seasonal, Seasonal Trend overview: Formula

Explanation of the Variables


Forecast Accuracy Measurements

Mean Absolute Deviation (MAD)Mean Absolute Deviation (MAD)Mean Absolute Deviation (MAD)Mean Absolute Deviation (MAD)

Error Total (ET)Error Total (ET)Error Total (ET)Error Total (ET)

Mean Absolute Percentage Error (MAPE)Mean Absolute Percentage Error (MAPE)Mean Absolute Percentage Error (MAPE)Mean Absolute Percentage Error (MAPE)

Mean Square Error (MSE)Mean Square Error (MSE)Mean Square Error (MSE)Mean Square Error (MSE)

Square Root of the Mean Squared Error (RMSE)Square Root of the Mean Squared Error (RMSE)Square Root of the Mean Squared Error (RMSE)Square Root of the Mean Squared Error (RMSE)

Mean Percentage Error (MPE)Mean Percentage Error (MPE)Mean Percentage Error (MPE)Mean Percentage Error (MPE)


Mean Absolute Deviation (MAD)

The Mean Absolute Deviation is calculated by the absolute differThe Mean Absolute Deviation is calculated by the absolute differThe Mean Absolute Deviation is calculated by the absolute differThe Mean Absolute Deviation is calculated by the absolute difference ence ence ence between the exbetween the exbetween the exbetween the ex----post forecast and the actual value.post forecast and the actual value.post forecast and the actual value.post forecast and the actual value.

The Mean Absolute Deviation is the error calculation used to The Mean Absolute Deviation is the error calculation used to The Mean Absolute Deviation is the error calculation used to The Mean Absolute Deviation is the error calculation used to determine the best fit in the Auto Model 2 Forecasting Model. determine the best fit in the Auto Model 2 Forecasting Model. determine the best fit in the Auto Model 2 Forecasting Model. determine the best fit in the Auto Model 2 Forecasting Model.


Error Total (ET)

The Error Total is the sum of the difference between the Actual The Error Total is the sum of the difference between the Actual The Error Total is the sum of the difference between the Actual The Error Total is the sum of the difference between the Actual and and and and the Exthe Exthe Exthe Ex----post forecast.post forecast.post forecast.post forecast.


Mean Square Error (MSE)

The Mean Square Error is the Average of the sum of the square ofThe Mean Square Error is the Average of the sum of the square ofThe Mean Square Error is the Average of the sum of the square ofThe Mean Square Error is the Average of the sum of the square ofeach difference between the exeach difference between the exeach difference between the exeach difference between the ex----post forecast and the Actual value.post forecast and the Actual value.post forecast and the Actual value.post forecast and the Actual value.


Mean Absolute Percentage Error (MAPE)

The Mean Absolute Percentage Error is the Average Absolute PerceThe Mean Absolute Percentage Error is the Average Absolute PerceThe Mean Absolute Percentage Error is the Average Absolute PerceThe Mean Absolute Percentage Error is the Average Absolute Percent nt nt nt Error between the Actual Values and their corresponding ExError between the Actual Values and their corresponding ExError between the Actual Values and their corresponding ExError between the Actual Values and their corresponding Ex----post post post post Forecast Values.Forecast Values.Forecast Values.Forecast Values.


Root mean Square Error (RMSE)

The Root Mean Square Error is the Square root of the average of The Root Mean Square Error is the Square root of the average of The Root Mean Square Error is the Square root of the average of The Root Mean Square Error is the Square root of the average of the the the the sum of the squared difference between the Actual and Exsum of the squared difference between the Actual and Exsum of the squared difference between the Actual and Exsum of the squared difference between the Actual and Ex----post post post post Forecast value.Forecast value.Forecast value.Forecast value.


Mean Percentage Error (MPE)Mean Percentage Error (MPE)Mean Percentage Error (MPE)Mean Percentage Error (MPE)

The Mean Percentage Error is the average percentage error of theThe Mean Percentage Error is the average percentage error of theThe Mean Percentage Error is the average percentage error of theThe Mean Percentage Error is the average percentage error of thedifference between the Exdifference between the Exdifference between the Exdifference between the Ex----post Forecast and the Actual values.post Forecast and the Actual values.post Forecast and the Actual values.post Forecast and the Actual values.


Thank You

Infodays-04 Fcst Models - Archive...historical periods Weighted Moving Average (14) Weighted Moving...

Documents

Transcript of Infodays-04 Fcst Models - Archive...historical periods Weighted Moving Average (14) Weighted Moving...