I want the model to be able to select from a different scope of columns or range of tables. How do I do that?

Answer:

In the One AI query builder, from the “Which Core Features do you want to use to predict (your target variable)?” step, you can select from the following scope options, which can be edited later as well:

• None
  • Only pulls in the unique identifier, allowing the model creator to manually select the columns they want the model to try
  • This is helpful when you only want the model to try a specific set of columns or attributes that doesn't fit neatly within the available scopes. 
• Narrow
  • Examines everything from the table from which your population is being pulled (often the Employee or Employee Event table)
  • This is a good place to start as it is less complicated and produces more easily understood models, EDA reports, and results summaries
• Balanced
  • Pulls tables that are directly joined (within 1 join) to the table where the population is being pulled from
  • An advantage of using the Balanced scope is that it incorporates user friendly dimension labels as opposed to the ID values that fact tables often contain for various attributes
• Broad
  • Pulls tables that are directly or indirectly joined to the table where the population is being pulled from
  • We do not recommend immediately using the Broad scope, but rather trying after you’ve tried Balanced and Narrow without getting desired results as it can produce complicated models that can be tricky to interpret

Available scopes for machine learning models in the One AI query builder