Configure AI Attributes

Learn how to configure the attributes associated with AI calculations, such as for AI insights, find similar, and explore this.

To set specific attributes for features like find similar, explore this, and AI insights, you can configure the attributes in Studio.

Note: For AI insight dynamic templates, attributes configurations are only used when the following requirements are met:

  • The function requires a set of properties or dimensions.
  • The template author did not provide a specific property or dimension.

For more information, see Create a Dynamic Template.

Configuration options

The following table defines the attributes you can configure.

Configuration Description Features
Properties

The default properties used in AI calculations. For example, this lists the attributes used to find a similar group in the Compare room or for the relatedGroup function in a dynamic template.

Note: The first 10 properties in the list are used to find similar groups in Compare. No other attributes are considered. For more information, see Use the Comparison Grid.
  • AI insights
  • Find similar
Change Properties The properties related to change that are used by default for AI calculations. For example, this lists the properties used in the similarChangeProperties function in a dynamic template.
  • AI insights
Grouping Dimensions

The dimensions that are used by default for AI calculations. For example, this lists the potential dimensions to explore a visual item or for the outlierGroups function in a dynamic template.

Note: The first 10 properties in the list are used to select a group by when exploring a visual item. No other group bys are considered. For more information, see Explore a Visual Item.
  • AI insights
  • Explore this

Note: The first 10 attributes in each population default are considered when performing AI calculations. However, this does not mean that the first property in the list will always be used. This is because Visier applies algorithms to find the best property to use in the context of your work.

Example: Exploring a Breakdown visualization

Let’s say you’re looking at a Breakdown of Headcount by Organization Hierarchy and you want to explore the Product organization. Your Grouping Dimensions population default is configured as follows:

  1. Job Family
  2. Job Name
  3. Gender

Visier’s algorithms may determine that Gender is a more meaningful dimension to group by than Job Family. When you click “Explore Product”, the visualization will then change to a Breakdown of Headcount by Gender in Organization Hierarchy > Product.

Configure AI attributes

To set the AI configuration:

  1. In a project, on the navigation bar, click Model > Analytic Objects.

  2. Select a subject, and then click AI Configuration.
  3. In the Configuration panel, select one of the following:
    • Properties
    • Change Properties
    • Grouping Dimensions
  4. Click Manage Properties or Add Dimension.

    Caution: Only the first 10 attributes in the list are used. If you exceed 10 attributes, the excess attributes will not be used in population calculations.

  5. Click Reorder to change the order of the properties from most to least important.

    Note: The first property in the list is weighted the most heavily in our calculations.

  6. When finished, publish the project to production. For more information, see Publish Project Changes.

Example: Find similar

Let’s say that you set the following Properties:

  1. Job Family
  2. Job Name
  3. Manager Status
  4. Time Since Promotion
  5. Gender

From this list, Visier will find similar groups to an individual using those five attributes, with Job Family weighted the heaviest.

Example: AI insights formula

Let’s say that you set the following Properties:

  1. Performance Rating
  2. Job Name
  3. Manager Status
  4. Time Since Promotion

From this list, Visier determines that Performance Rating is the most important property to consider when performing AI calculations.

The following insight formula will search for related groups based on the properties configured in the list above.

triggerPopulation.relatedGroup(scopePopulation, 2)

Because there are no properties identified in the formula, the formula will use the list of properties to find related groups.

The following formula is the equivalent of the above formula, except with the properties specified directly in the formula.

triggerPopulation.relatedGroup(scopePopulation, 2, propertiesToConsider = ["Employee.Performance_Rating", "Employee.Job_Name", "Employee.Manager_Status", "Employee.Time_Since_Promotion"])

The above formula would not reference the configured list of properties because the properties are identified in the formula itself.