Handling Decimal Values and Negative Values in Auto Predict and Predictive Planning

Auto Predict and Predictive Planning handle decimal values and negative values so that they are appropriately applied to members in predictions.

Handling Decimal Values

Auto Predict and Predictive Planning automatically handle decimal values in predictions based on the account and whether it’s valid for the account to have decimal predicted values. When needed, the prediction engine uses GenAI to decide how to process values. This ensures that decimal values are handled properly for all predictions without any intervention required from users.

Not all accounts are valid to have decimal predicted values, for example, accounts such as Headcount, New Hires, or Number of Stores. Auto Predict and Predictive Planning automatically check the account to determine whether predictions can have decimal values, and then apply adjustments during the post prediction process if needed.

GenAI must be enabled for this feature to work. If GenAI is not enabled, then no post prediction processing is done, and decimal values can show up for accounts or measures such as head count, even though historical data has no decimal values. See Enabling AI in FreeForm.

Once a prediction is generated, if decimal values are found, Auto Predict and Predictive Planning use this process for determining how to handle the decimal values:

  1. Check historical data and see how it is handled. If historical data has decimal values, then decimal values are valid for predictions, and no further are steps taken.

    If historical data does not have decimal values, move to the next step.

  2. Check the data type. If the data type is currency or percentage, it is valid to have decimals for the prediction, and no further steps are taken.

    If the data type is not currency or percentage, move to the next step.

  3. Use GenAI and LLM to validate the account details such as account type, alias name, and account name and determine whether it’s valid to have decimal prediction values.

  4. Apply adjustments during post prediction processing.

    For an account that can’t have decimal predictions, values are rounded off to the nearest integer.

Auto Predict and Predictive Planning first process negative predictions if negative values are found, and then process predictions with decimal values.

Handling Negative Prediction Values

Auto Predict and Predictive Planning handle negative values in predictions based on the account and whether it’s valid for the account to have negative predicted values. Optionally, the prediction engine can use GenAI when needed to decide how to process values. This ensures that negative values are handled properly for predictions.

Not all accounts are valid to have negative predicted values, for example, accounts such as headcount or units sold. Other accounts, such as marketing expense, could have negative predicted values.

Auto Predict and Predictive Planning check the account to determine whether predictions can have negative values, and then apply adjustments during the post prediction process if needed.

Optionally, you can use GenAI if needed to help determine whether predictions can have negative values. Using this option requires you to enable GenAI. See Enabling AI in FreeForm.

To select options for handling negative predictions:

  1. For Auto Predict:
    1. From the Home page, click IPM, and then click Configure.
    2. Select an Auto Predict job or create a new one.
    3. On the Configure page of the IPM Configuration wizard, click Show Advanced Options, and in the Prediction Constraints area, select Adjust Invalid Negative Predictions. With this option selected, the prediction engine checks historical data to determine how to handle negative prediction values.
    4. Optionally, select Detect non-negative measures using GenAI to use GenAI if needed to determine how to handle negative prediction values.

      GenAI must be enabled for Detect non-negative measures using GenAI to work. If GenAI is not enabled, only historical values are checked for negative prediction values.

  2. For Predictive Planning:
    1. After you run a Predictive Planning prediction on a form, in the results area, click Settings icon(Settings) and then click Prediction Options.
    2. In the Prediction Constraints area, select Adjust Invalid Negative Predictions. By default this option is off.
    3. Optionally, select Detect non-negative measures using GenAI to use GenAI if needed to determine how to handle negative prediction values.

Once a prediction is generated, if negative predicted values are found, Auto Predict and Predictive Planning use this process for determining how to handle the negative values:

  1. If Adjust Invalid Negative Predictions is selected: Check historical data and see how it is handled. If historical data has negative values, then negative values are valid for predictions, and no further are steps taken.

    If historical data does not have negative values, and Detect non-negative measures using GenAI is selected, move to the next step.

    If historical data does not have negative values, and Detect non-negative measures using GenAI is not selected, then negative predictions are adjusted to zero.

  2. If Detect non-negative measures using GenAI is selected, use GenAI and LLM to validate the account details such as account name, alias name, account type, hierarchy type, parent name and determine whether it’s valid to have negative predicted values. (If GenAI is not enabled, only historical values are checked for negative prediction values.)

  3. Apply adjustments during post prediction processing.

    For an account that can’t have negative predictions, the negative values are adjusted to zero.

Auto Predict and Predictive Planning first process negative predictions if negative values are found, and then process predictions with decimal values.

The downloadable Auto Predict report shows details about adjusted negative predictions. See the columns: Adjusted Negative Predictions Count and Updated Negative Predictions Period.