Measure mapping options for ATT forecast engine
You can select the options on the Measure Mapping tab on the Forecast Engines page for the ATT forecast engine.
This table displays the measure mapping options:
Measure  Description 

History  The measure that contains the history values against which the forecast data is generated. 
Period Weighting  The measure that contains the period weighting data. The CalculatePeriodWeighting macro command is used to populate the weighting measure. The period weighting data is used to normalize the history and forecast data. This is a required field. 
History Mask  The measure that contains the history mask data. This measure is used to indicate the history periods that are not used to generate a forecast. This is a required field. 
Forecast  The measure to which the statistical forecast data generated by the forecast engine is transferred. 
Standard Deviation  The measure to which the statistical standard deviation data generated by the forecast engine is transferred. 
Level  The measure to which the statistical level data generated by the forecast engine is transferred. 
Growth  The measure to which the statistical growth (trend) data generated by the forecast engine is transferred. 
Online Model Fit  The measure that must be used to store the online forecast model, using the history periods, generated by the forecast engine. 
Short History Exception  The measure to which the short history exception data for the specified item or location combinations is transferred. The forecast is not generated for the item or location combination for this exception. 
Tracking Signal Exceptions  The measure to which the tracking signal
exception data, for the generated forecast, is transferred. The exception is
returned only when the default algorithm is set to HoltWinters, Least Squares,
or Moving Average. In case, the forecast calls Best (picking algorithm), the
exception is returned only when the generated forecast is based on
HoltWinters, Least Squares, or Moving Average. The measure that receives the resulting tracking signal exceptions for the generated forecast. The periods > 0 have a tracking signal exception. The value 1 is also written to pconst to indicate that the item/location has at least one tracking signal exception within the tracking exception horizon. 
Outliers  The measure to which the outlier exception
data, for the generated forecast, is transferred. The exception is returned
only when the default algorithm is set to Least Squares. In case, the forecast
calls Best (picking algorithm), the exception is returned only when the
generated forecast is based on Least Squares. The measure that receives the resulting outlier exceptions for the generated forecast. The history period > 0 have an outlier and are masked from the model fit process. The value 1 is written to pconst to indicate the item/location has at least one outlier within the outlier exception horizon. 
Step Change Exceptions  The measure to which the step change
exception data, for the generated forecast, is transferred. The exception is
returned only when the algorithm is set to HoltWinters. In case, the forecast
calls Best (picking algorithm), the exception is returned only when the
generated forecast is based on HoldWinters. The measure that receives the resulting step change exceptions for the generated forecast. The periods > 0 have step change exception indicating the first period where a number of contiguous outliers are observed. The value 1 is also written to pconst to indicate the item/location has at least one step change exception within the step change exception horizon. 
Algorithm  The measure which stores the algorithm used
by the forecast engine. This is mainly applicable when calling the engine using
Best method (picking algorithm). The return value can be one of these:
Note: The data is stored as an enumeration in SCP as it is not
possible to store measures of Text type.

Auto Correlation Coefficient  The measure to which the auto correlation (defines the measure of relationship between the two stages of forecasting) coefficient value generated by the forecast engine must be transferred. This value is zero, if two stage forecasting is not used. 
Degrees of Freedom  The measure to which the degrees of freedom
(the amount of independent information that must be included to estimate a
parameter in the final calculation) value generated by the forecast engine must
be transferred. This value indicates the number of data points that must be
used in the model fit that does not include the number of data points in the
startup window. Note: This is not applicable for ARIMA
algorithm.

History Trend Damping Factor  The measure to which the history damping factor (a corrective factor used to minimize the instability in the data collected in the exponential smoothing process) used by the forecast engine must be transferred. This value is a number that is multiplied by the trend value (growth rate) for the calculation of each forecast value at the specified time. 
Information Criterion  The measure to which the information criterion used by the forecast engine must be transferred. This measure is used to compare the algorithm models. 
Level Smoothing Coefficient  The measure to which the level, that is the component (smoothing constant) in the exponential smoothing process used to generate forecasts, of the smoothing coefficient (used by the forecast engine) must be transferred. The level smoothing coefficient (alpha) is selected by fitModelPickingAlgorithm(), when the forecast engine type is HoltWinters, or when the Optimize parameter is set to TRUE, and the algorithm is HoltWinters. 
LjungBox P Value  The measure to which the LjungBox P value (statistical test value) must be transferred. This value is used to test whether the residuals (history point minus one step ahead forecast) of a model are significant. 
LjungBox P Value Significant  The measure to which the value of the LjungBox P Value Significant flag must be transferred. This value determines whether the LjungBox test is at a significant level. 
Model Fitting History  The measure to which the history data used by the forecast engine in the model fitting process must be transferred. 
Model Form  The measure to which the model form used by
the selected forecast engine technique must be transferred. The return value
can be one of these:

Moving Point Average  The measure to which the number of moving average points used by the forecast engine must be transferred. When the forecast engine technique is Moving Average, this indicates the value selected by the fitModelPickingAlgorithm(). 
Obsolescence Exception  The measure to which the obsolescence (indicates the item has no sales, zero forecast, in the specified horizon) exception data from the forecast engine is transferred. This indicates that the forecasted level is negative within the cycle period forecast horizon. This value is the projected level after the corresponding growth is applied to each level of the future period in the cycle period horizon. The step changes or growth dampening is not included in this calculation. 
Seasonal Smoothing Coefficient  The measure to which the seasonal (type of periodical component used in the smoothing technique, time series analysis, that is used to generate forecasts) smoothing coefficient used by the forecast engine must be transferred. The seasonal smoothing coefficient (gamma) is selected by fitModelPickingAlgorithm(), when the forecast engine is of the type, HoltWinters, or when the Optimize parameter is set to TRUE, and the algorithm is HoltWinters. 
Seasonal Type  The measure to which the seasonal (type of
periodical component used in smoothing technique) type used by the forecast
engine must be transferred. The return value can be one of the following:

Seasonal Indices  Indicates the measure that receives the seasonal indices for the model, if applicable. This represents the extent to which the average for a particular period tends to be above (or below) the expected value. The type of seasonal indices that are generated are displayed in the Seasonal Type and Model Form. Additive seasonal models are absolute values which are added to the levels plus the growth. Multiplicative models are multipliers wherein the level plus the growth are multiplied by the seasonal index. Generates 'n' values, where n represents the Periodicity, or observations for the year. For example, if the monthly data is considered there are 12 separate seasonal indices, that is, one for each month. 
SMP Chance of Event  The measure to which the SMP (Slow Moving Part) chance of event (mean interarrival time) of nonzero demand values from the Crostons algorithm must be transferred. This parameter is not populated (returns 0) if the item is not an SMP. 
Tracking Signal Type  The measure to which the tracking signal (a
signal that monitors any forecasts generated to compare with the actuals and
generates a warning when unexpected departures are identified in the forecast
results) type used by the selected forecast engine type, must be transferred.
The return value can be one of these:

Trend Smoothing Coefficient  The measure to which the trend (type of component used in the smoothing technique, time series analysis, that is used to generate forecasts) smoothing coefficient used by the forecast engine must be transferred. The trend smoothing coefficient (beta) is selected by fitModelPickingAlgorithm(), when the forecast engine type is HoltWinters, or when the Optimize parameter is set to TRUE, and the algorithm is HoltWinters. 
Combining Type  Indicates the Combining type used by the
engine. Possible return values:

Probabilistic Standard Deviation  Indicates the measure that receives the Probabilistic Standard Deviation for the generated forecast. The probabilistic standard deviation is a back calculation based on the calculated Prediction Intervals. The result is returned in parallel to the classical Standard Deviation. 
Lower Prediction Interval  Indicates the measure that receives the lower prediction
interval for the generated forecast. This measure is mapped and used along with
the Upper Prediction Interval, though the application supports mapping only one
measure or None. Prediction intervals (PIs) are upper and lower limits around the forecast mean which indicate the range of values that the actual value considers with a specified probability, that means the real value is likely to be within limits defined by the upper and lower prediction intervals. For example, if it is assumed 95% Prediction Intervals (PIs), the observation associated with that first forecast period to fall within the PIs 95% of the time. Note: This measure is not
applicable for CROSTONS or EVENTSONLY algorithms.

Upper Prediction Interval 
Indicates the measure that receives the upper prediction interval for the generated forecast. This measure is mapped and used along with the Lower Prediction Interval,though the application supports mapping only one measure or None. Prediction intervals (PIs) are upper and lower limits around the forecast mean which indicate the range of values that the actual value considers with a specified probability, that means the real value is likely to be within limits defined by the upper and lower prediction intervals. For example, if it is assumed 95% Prediction Intervals (PIs), the observation associated with that first forecast period to fall within the PIs 95% of the time. Note: This measure is not applicable for CROSTONS or
EVENTSONLY algorithms.

Additional measure mapping options:
See Measure mapping for the Multiple Regression algorithm.
This table displays the Event Modeling measures:
Measure  Description 

Average Event Size  The measure to which the average event size,
in item units, is added when event modeling is executed. This is applicable
when
Perform Event Modeling = TRUE .

Events  The measure to which the event profile data
must be added. An event profile contains the historic and future ratios for
periods defined by the user. This is mandatory when
Perform Event Modeling = TRUE .
The event profile Indicates the history and forecast periods that store the 'event magnitude', which represents the number of events that are included in a history period (this is not the same as the event size). This data is transferred to the forecast engine for event modeling, which is used to create the average event size, baseline history, and forecast. 