Anda di halaman 1dari 5

Forecast Strategy Forecast Method Type

Forecast with constant model Constant

First-order exponential smoothing Constant

Constant model with automatic alpha Constant


adaptation (first-order)

Moving average Constant

Weighted moving average Constant

Forecast with trend model Trend


Trend model with automatic alpha adaptation Trend
(second-order)

Forecast with seasonal model Seasonal

Seasonal model based on Winters' method Seasonal

Seasonal linear regression Seasonal

Forecast with automatic model selection. Test


for constant, trend, seasonal, and seasonal Automatic
trend (model selection procedure 1).
Test for trend (model selection procedure 1) Automatic

Test for season (model selection procedure 1) Automatic

Test for trend and season (model selection Automatic


procedure 1)

Seasonal model and test for trend (model Manual


selection procedure 1)

Trend model and test for seasonal pattern Manual


(model selection procedure 1)

Historical data used as forecast data Copy History

Manual forecast Manual

Croston Method Croston

Simple linear regression Linear Regression


System Action

Uses first-order exponential smoothing. The older the time series values, the less important they are for the
calculation of the forecast. The present forecast error is taken into account in subsequent forecasts.

It uses a smoothing constant (alpha) to which a value between 0 and 1 is assigned. The larger its value (closer to
1), the more weight it assigns to recent sales history. A large alpha (.8) is comparable to using a small number of
time periods (n) in a moving average model. A small n allows greater emphasis to be placed on recent periods.
Conversely, a small alpha (.1) is similar to using a large number of time periods in the moving average, because
the impact of recent data is lessened.

Uses first-order exponential smoothing and adapts the alpha factor.

The system calculates the average of the values in the historical time horizon as defined in the master forecast
profile. This average for n periods of history is the forecast result for each period in the forecast horizon; that is,
the forecast is the same in every period.

The system weights each time series value with a weighting factor. For example, you can define the factors so
that recent data is weighted more heavily than older data. You define the weighting factor in a diagnosis group.

A gradual upward or downward slope or Trend is captured and taken into consideration.

Uses second-order exponential smoothing and adapts the alpha factor.

A cyclic pattern with crests and troughs is expected to be captured out of the Sales pattern with a specified
Season or Period.

Winters' method employs a level component, a trend component, and a seasonal component at each period. It
uses three weights, or smoothing parameters, to update the components at each period. Initial values for the
level and trend components are obtained from a linear regression on time. Initial values for the seasonal
component are obtained from a indicator variable regression using detrended data.

Calculates seasonal indexes, removes the seasonal influence from the data, performs linear regression, and
reapplies the seasonal influence to the calculated linear regression line. See also Seasonal Linear Regression.

Choose this strategy if you have no knowledge of the patterns in your historical data.
The system tests the historical data for constant, trend, seasonal, and seasonal trend patterns. The system
applies the model that corresponds most closely to the pattern detected. If no regular pattern is detected, the
system runs the forecast as if the data revealed a constant pattern.
In this process, the alpha, beta, and gamma factors are determined as follows:
· The smoothing factors are taken from the univariate profile.
· The settings in the demand planning desktop are used if these are different to the ones in the univariate
profile.
· If you have made no settings either in the univariate profile or on the demand planning desktop, the default
factors of 0.3 are used.
Choose this strategy if you think that there is a trend pattern in your historical data, and if you know that there is
no other pattern.
The system subjects the historical values to a regression analysis and checks to see whether there is a significant
trend pattern. If not, the system runs the forecast as if the data revealed a constant pattern.

Choose this strategy if you think that there is a seasonal pattern in your historical data, and if you know that
there is no other pattern.
The system clears the historical values of any possible trends and carries out an autocorrelation test. If no
seasonal pattern is detected, the system runs the forecast as if the data revealed a constant pattern.

Choose this strategy if you think that there is a seasonal and/or a trend pattern in your historical data.
The system subjects the historical values to a regression analysis and checks to see whether there is a significant
trend pattern. It also clears the historical values of any possible trends and carries out an autocorrelation test to
see whether there is a significant seasonal pattern. If a seasonal and/or trend pattern is detected, a trend
model, seasonal model, or seasonal trend model is used. If no regular pattern is detected, the system runs the
forecast as if the data revealed a constant pattern.
Choose this strategy if you think that there is a trend pattern in your historical data, and if you know that there is
a seasonal pattern.

The system subjects the historical values to a regression analysis and checks to see whether there is a significant
trend pattern. If there is, a seasonal trend model is used. Otherwise, a seasonal model is used.

The alpha,
Choose thisbeta, and ifgamma
strategy factor
you think thatis there
also determined aspattern
is a seasonal described under
in your strategydata,
historical 50. and if you know that
there is a trend pattern.

The system clears the historical values of any possible trends and carries out an autocorrelation test. If the test is
positive, a seasonal trend model is used. Otherwise, a trend model is used.

The alpha, beta, and gamma factor is also determined as described under strategy 50.

Choose this strategy if demand does not change at all and you want to opt for the least performance- or work-
intensive strategy. No forecast is calculated. Instead, the historical data from the previous year is copied.

Choose this strategy if you wish to set the basic value, trend value, and/or seasonal indexes yourself. Forecast is
done with manual intervention by the expert.
Choose this method if demand is sporadic.
The system calculates a line of best fit for the equation y = a + bx, where a and b are constants. The ordinary
least squares method is used.
Parameters used along with the Forecasting Models

Alpha - It typically represents the weightage given to the recent sales history
(actual value). It's value varies from 0 to 1. If it's closer to 1, you give more
weight to recent value.

Beta - This parameter helps to represent the trend in the demand pattern. It's
a smoothing factor to trends i.e. closer to 1 gives more weight to the recent
trend.

Gamma - This factor helps to adjust the forecast by applying seasonal


percentage. It's again a smoothing factor to seasonal index. If the value is
closer to 1, more weight will be given to recent seasonal periods.

Seasonal index of a period = avg sales per period / avg sales of entire
horizon
Sigma - Outlier correction or Tolerance