Skip to main content

​
Overview

This guide walks you through creating Classification Cohorts from a trained Classification Model. Once your model has finished training, you can create cohorts for each label at different confidence thresholds.
Prerequisites:
  • A trained Classification Model (training takes approximately 12 hours)
  • Access to the Permutive Dashboard with cohort creation permissions

​
Understanding the Confidence Curve

Once the model has been trained, you will see a curve that shows the expected reach for each point along the confidence scale. Classification cohort confidence curve The confidence measures how sure the model is about a given prediction. While the reach of a classification model is the entire audience, the model will be less certain about some users than others. Use the confidence threshold to restrict your model to only classify users it is more certain about.

​
Steps

1

Select confidence threshold

Move the slider to select a minimum confidence threshold.
2

Choose dataset class

Click on one of the Dataset classes buttons to create a cohort for each of the labels in the model.
3

Confirm creation

A modal will appear asking for confirmation to create the Classification Cohort.
Confirm creation of a classification cohort
Multiple cohorts: You can create multiple Classification Cohorts from the same model using different combinations of seed labels and confidence thresholds.

​
Enabling Activations

Once you create the Classification Cohort, you will find it on the Modeled Cohorts page. Here you can also enable activations for the cohort to push to ad servers such as Google Ad Manager and Xandr. Activations for a created classification cohort

​
Next Steps

Viewing Classification Cohorts

Manage and filter your cohorts

Back to Classification Cohorts

Return to product overview