Skip to main content

Intelligent Forecast Dataview [v26.1] 

The Intelligent Forecast Dataview is an AI-powered forecasting feature that generates data-driven predictions using the Forecast service. You can create an Intelligent Forecast Dataview to define forecast parameters based on historical data, save the configuration, and use the dataview in templates to produce forecasted results.

The Intelligent Forecast Dataview uses historical data from an existing SQL Query or Autogenerated dataview and passes it to the Forecast service, which applies machine learning algorithms to generate forecast values for the specified time period.

Prerequisites

  • The Intelligent Forecast feature toggle is enabled.

  • The Forecast service endpoint is configured under Application Settings.

  • At least one SQL Query or Autogenerated dataview exists with historical data containing measure fields and time-based data.

  • Active time units are configured in the system.

  • Suggested historical data for at least 3–5 years for optimum results.

Configure the Forecast Service Endpoint

Before creating an Intelligent Forecast Dataview, an administrator must configure the Forecast service endpoint.

  1. Navigate to Application Settings.

  2. Locate the Forecast service endpoint for Intelligent Forecast field.

  3. Enter the Forecast service URL for the appropriate Platform environment (development, staging, or production).

  4. Click Save.

Note: The system authenticates with the Forecast service using the Platform token. If the connection fails, it returns an error. Connection errors are logged with meaningful messages.

 

Add an Intelligent Forecast Dataview[v26.2] 

To create a new Intelligent Forecast Dataview:

  1. Navigate to the Dataviews page.

  2. Click Add and select Intelligent Forecast Dataview.

    The Intelligent Forecast Dataview editor opens as a full-page view, replacing the previous modal dialog.

  3. Complete the following fields. Fields are grouped into sections as displayed in the editor.

    Field Description
    General
    ID Enter a unique identifier for the dataview. Alphanumeric values only. The Enter key is recorded as an underscore. No other special characters are allowed.
    Description Enter a description for the dataview.
    Data Source
    Source Data View Select a historical dataview for the forecast from the dropdown. Lists all available dataviews that can be used as the source. Mandatory.
    Value Field Select the measure field from the historical dataview to use for forecast values. Mandatory.
    Time Unit Select the time column from the dropdown. Mandatory.
    Training Period
    Start Period Enter the start period of the historical data (for example, 202101). Mandatory.
    End Period Enter the end period of the historical data (for example, 202512). Mandatory.
    Seasonal Period Select whether the data follows a repeating seasonal pattern. When a seasonal period is selected, the Forecast service detects recurring cycles in historical data and uses them to produce more accurate forecasts — for example, modelling year-end expense peaks for monthly financial data. Options: None (default), Monthly – Annual (12), Quarterly – Annual (4), Weekly – Annual (52). A help icon next to the field provides guidance on selecting the correct option. Requires at least 24 months of historical data per account for seasonal detection to apply.
    Forecast Settings
    Forecast Coverage Algorithm Select the forecasting algorithm from the dropdown. Options: Automatic (default), AutoARIMA, AutoTHETA, AutoETS. See Forecast Algorithms below for guidance on each option.
    Enable lower and upper bound values Select this checkbox to include lower and upper bound values in the forecast results. Enabled by default.

  4. Click Save.

    The system saves the dataview configuration and displays a success message. The dataview appears in the dataviews list on the Dataviews page.

Note: The Save button remains disabled until all mandatory fields are completed. If the operation fails, the system displays a failure message.

Once saved, the Intelligent Forecast Dataview generates prediction fields based on the selected measure. If lower and upper bound values are enabled, the corresponding limit fields are included.

Seasonal Period Options [v26.2] 

The Seasonal Period field controls whether the Forecast service looks for repeating patterns in the historical data. Selecting an option other than None instructs the service to detect and model seasonal cycles before generating forecast values.

Option Description
None (default) No seasonal pattern is applied. The Forecast service distributes values evenly based on trend only. Use this for flat or non-seasonal data. Existing Dataviews that do not yet have this field set retain this behaviour by default.
Monthly – Annual (12) Applies a 12-period seasonal cycle. Use this for monthly data with repeating annual patterns — for example, year-end peaks or consistent dips at the start of the year in financial accounts.
Quarterly – Annual (4) Applies a 4-period seasonal cycle. Use this when the time column represents quarterly data with annual seasonal patterns.
Weekly – Annual (52) Applies a 52-period seasonal cycle. Use this for weekly data with repeating annual patterns.

Note: Seasonal detection requires at least 24 months of historical data per account. Accounts with less history automatically use non-seasonal forecasting regardless of the Seasonal Period setting selected.

Forecast Algorithms

The following algorithms are available when configuring an Intelligent Forecast Dataview:

Algorithm Description

Automatic (Recommended)

Runs all three algorithms, compares accuracy scores, and selects the best fit automatically. Recommended for most use cases.

AutoARIMA

Detects trends by studying how recent values relate to previous values. Learns from past errors and adjusts future predictions. Works best when values grow or decline consistently month over month.

AutoETS

Learns seasonal rhythm—months that are always high or always low. Gives more weight to recent history for more responsive forecasts. Requires at least 24 months of data to detect seasonality reliably.

AutoTheta

Splits the data into long-term direction and short-term movement, forecasts each independently, then combines them. Most resilient when data is volatile or unpredictable.

Edit an Intelligent Forecast Dataview

To edit a saved Intelligent Forecast Dataview:

  1. Navigate to the Dataviews page.

  2. Click the Intelligent Forecast Dataview you want to edit.

    The editor opens in full-page view, displaying the previously configured parameters.

    Note: The ID field is not editable once the dataview is saved..

  3. Update the required fields.

  4. Click Save to save the changes, or click Cancel to discard them.

Use an Intelligent Forecast Dataview in a Template

To generate forecast data using the Intelligent Forecast Dataview:

  1. Open a template in Template Design.

  2. Add the Intelligent Forecast Dataview as the data source.

  3. Select the forecast field to use as a column.

    Predicted field names follow the format:

    [measure]_pred, [measure]_pred_upper, [measure]_pred_lower.

    For example, if the measure field is amount, the predicted column is amount_pred.

  4. Execute the template in preview mode.

    The system extracts the saved forecast parameters, transforms the historical data into the format required by the Forecast service, and returns the forecasted data for use in the template.

  5. To inspect the execution, open Trace to view the query, the data generated, and the request sent to the Forecast service via the Get Predictive Data button.

Was this article helpful?

We're sorry to hear that.