In a binary classification machine learning model, what is "the positive label"?


The most common type of machine learning model in One AI is a binary classification.  A binary classification predicts whether or not an outcome is likely to occur for each individual record.  The model actually makes predictions on both sides of the coin though.  Using attrition as an example, the model predicts both likelihood of terminating and also likelihood of not terminating.  Since the likelihood of not terminating is just the inverse of the likelihood of terminating, it’s more useful to just review one outcome.  The outcome you want to focus on, in this case likelihood of terminating, is the positive label. The the positive label typically refers to the outcome or class of interest that the model is designed to predict.

The positive label is identified on the Results Summary Report in the upper right section (One AI >Runs > Run Label > Results Summary)

Positive Label shown on the Results Summary Report

The positive label is automatically chosen by One AI as the last value when ordered ascending, but can be manually configured in the One AI query builder in the override values step (Augmentations >Edit > Configure One AI Recipe > Override Values step). 

Configuring the Positive Label from the One AI query builder

Was this article helpful?

0 out of 0 found this helpful



Please sign in to leave a comment.