Understand category classification model performance

  • Article

After each training, AI Builder uses the test dataset to evaluate the quality and accuracy of your AI model. A summary page for your model shows your model training results, including a Performance score.

AI Builder calculates your model's performance score based on the precision and recall of the prediction results, as well as the F1 score:

  • Performance score: This score is calculated using precision, recall, and F1 scores. Performance score values are from 0 and 100. Generally, the higher the performance score, the better your model is.
  • Precision: The fraction of correct predictions among all the positive predictions.
  • Recall: The fraction of correct predictions among all true positive cases.

Quick test

You can also select Quick Test to assess the quality of the model. Just enter text that you want to tag. More information: Evaluate your model

Next step

Improve the performance of your category classification model