Setting Prediction Options

The prediction options specify data attributes, prediction methods, and other aspects of time-series analysis performed by Predictive Planning. The defaults are suitable for most predictions and should only be changed by those with some knowledge of time-series analysis.

To set prediction options:

1. Open the Settings dialog.
2. On Options, review and select from the following:

   • Data attributes group:

     • Select whether to detect seasonality (regular cycles of data) automatically (Automatic, the default) or manually (Manual). If you select Manual, specify the number of time periods per cycle For example if time periods are quarters with a yearly cycle, there would be 4 periods per cycle.

     • Select whether to Fill in missing values and Adjust outliers. These settings estimate missing data based on adjacent data and help to normalize unusual data.

       Note:

       Fill-in Missing Values uses interpolation to fill in gaps in the historical data. Clearing this option skips prediction calculation for members with gaps in their data.

       Adjust Outliers uses a special fitting algorithm to determine whether data points fall within a reasonable range compared to all the other data points for a member. Clearing this option still allows the prediction to proceed, although the prediction algorithm may be thrown off by the outlier data points.

   • Prediction methods group:

     • Select which time-series prediction methods to use: Nonseasonal (does not fit to cyclical data), Seasonal (fits to cyclical data), or ARIMA (both nonseasonal and seasonal using predefined statistical models). See Classic Time-series Forecasting and ARIMA Time-series Forecasting Methods for lists and details.

       Select all three, the default, unless you have a good reason to do otherwise.

     • Select an error measure to use in selecting the best method: RMSE, MAD, or MAPE (Time-series Forecasting Error Measures).

       Again, use the default, RMSE, unless you have a good reason to use another.

   • Prediction periods group:

     • Select whether to detect periods automatically, Select periods based on form, or manually, Manual. If you select Manual, specify the number of periods to predict. Generally, the number of prediction periods should be less than half the amount of actual data.

     • Select a Prediction interval, which defines a range around the base predicted value where the value has some probability of occurring; for example, the default (2.5% and 97.5%) means that there is a 95% probability that the predicted value will fall between the 2.5 percentile and the 97.5 percentile.

       Note:

       Prediction Interval determines the percentile range around the base case prediction that is used to represent the best and worst case predictions. For example, a 2.5% - 97.5% prediction interval estimates that 95% of the time the predicted value will actually occur between the lower and upper bounds; 5% of the time the value will lie outside of these bounds.

       These lower and upper percentile values are also used to indicate the worst and best case predicted values. For a Revenue-type account member, the worst and best cases are assigned to the lower and upper percentile values, respectively. For an Expense-type account member, the cases are reversed; the best case is associated with the lower bound (e.g. 2.5%) and the worst case is associated with the upper bound (e.g. 97%).

3. Optional: Set or reset defaults using one of the following selections:

   • Click Set Default to store settings on all tabs as application defaults.

   • Click OK to store settings on all tabs as individual defaults for only this form.

   • Click Reset at any time to restore the predefined defaults shipped with Predictive Planning or application defaults set with Set Default. This resets all tabs of the dialog.

     Note:

     For more information about defaults, see Application and Individual Form Defaults.

4. Optional: To leave the dialog without changing defaults, click Cancel.