Use the Advanced tab to set the scope of historical data, define a
confidence interval, define where you want the forecast values to appear, and set data spreading
options for writing forecast data into consolidated members. Some options on the
Advanced tab are not applicable to multivariate forecasts, as noted in the
option descriptions.
Some of these Advanced options are not available to all user roles. When applicable,
these restrictions are also noted in the option descriptions.
The following options are available on the Advanced tab.
- Seasonality
- Not applicable to multivariate forecasts. This option is disabled for consumer users.
- If your model is highly seasonal, consider using a univariate forecast.
- This option specifies the seasonality with which to build the model. Seasonality is when the
time series has a predictable cyclic variation. For example, during a holiday period each
year.
The default value is Auto-detect. Auto-detect
automatically detects seasonality by building multiple models with different seasonal periods and
choosing the best one.
- If you switch Auto-detect off, you can specify seasonality by entering a
non-negative integer, such as 0, 1, 2, 3 in the Enter seasonality interval
field.
- To specify a non-seasonal model, set the Enter seasonality interval to 0
or 1. A model with user specified seasonality is displayed only if the seasonal model is more
accurate than all of the non-seasonal models.
- Select scope of historical data used
-
This option is disabled for users with the analyst or consumer user role.
Select between the following two options:
- Use historical data in TM1 cube
- Use historical data in the Exploration
The default option is
Use historical data in the Exploration .
In this example, the leave members of Period hierarchy start at Jan 2018 and goes on
to Sept 2021. However, the exploration view starts at Jan 2019.
-
-
When the option Use historical data in the Exploration is selected, the
historical data that is used for the prediction starts from Jan 2019 and includes only the members
in the current Exploration view. As shown in the next image of the forecast preview, the
visualization starts at Jan 2019, which is the start of the Exploration view and also used as the
starting point for the history. Data for Jan 2018 - Dec 2018 is not included in the history of the
prediction in this case.
-
Conversely, when you use the option Use historical data in TM1 cube, the
leaf members of this Period hierarchy start at Jan 2018.
-
Based on this selection criteria, the forecast preview uses data from Jan 2018 onward to include
all members in the Period dimension as the historical data. The following forecast preview image
shows that the start point of the history is Jan 2018. Because more data is included as the history,
and that data contains a decline, the forecast results changes.
- Select confidence interval
- This option is disabled for consumer users.
- Not applicable to multivariate forecasts, which do not predict a range with high and low
values.
- The certainty with which the true value is expected to be within the specified range. You can
see corresponding confidence interval in a tooltip by hovering over any forecast value. The
confidence interval is displayed as higher and lower bounds.
- You can select three different confidence intervals: 90%, 95%, and 99%. The default is 95% and
the lower and higher bound define the range at which you can be 95% confident that the true value
lies within that range.
- Adjust outliers
- When this option is turned on, all detected outliers are corrected before performing the
forecast. You can preview the effect of this by selecting a row that contains an outlier and
clicking the Preview button. See Previewing with detected outliers for further information on how Adjust
outliers works.
- Ignore historical time periods
- This option is disabled for consumer users.
- Ignores a specified number of data points when building the model and computing the forecasts.
You can either ignore a set of data points at the end of a time series, or ignore individual data
points in different places in a time series. The ignored data points are interpolated. A linear
interpolation is used to populate the ignored values. Select the ignored periods in the nested
Ignore time periods pane. You can select periods either as individual points
or from a starting point onwards, until the start of the forecast.
Note: When you ignore historical time periods, the selected time periods in only time series are
ignored. Time periods are not ignored in external variable time series.
- By default, no historical data is ignored. If no missing values exist, then all of the
historical data is used and the first forecast point is after the last historical data point.
- Ignoring data period can be useful when the data is incomplete. For example, you might be doing
a forecast halfway through a month. Exclude this month from the forecast by setting
Ignored historical time periods either to that specific month, or to the last
data point if that month is your last.
- Spread forecast values
- Define the data spreading method that you want to use to spread data from consolidated cells.
- You can use the Proportional and Relative proportional data spreading methods. For more
information, see Apply data spreading to a forecast.
- Where do you want to save the predicted values
-
This option is disabled for users with the analyst or consumer user role.
This section lets you identify the dimension, hierarchy, and member where you would like to save
your forecast values. For example, you would typically forecast from the ‘Actual’ member of the
‘Version’ dimension into the ‘Forecast’ member of the ‘Version’ dimension.
For univariate forecasts, you can also specify the members where you want to save upper-bound and
lower-bound forecast values. If you want to save the upper-bound and lower-bound values for
consolidations, select the Save upper and lower bound for consolidations
option. This option is not available for multivariate forecasts, which do not generate upper-bound
and lower-bound forecast values. This option is not available to consumer users.