Category graph: Precision-Recall vs. Threshold

You can use the Precision-Recall vs. Threshold graph to determine category thresholds, or examine the performance of the selected category.

The Precision-Recall vs. Threshold graph displays the recall and precision rates for the selected category, based on test results.

• The light (blue) line represents the recall rate at different threshold settings
• The dark (maroon) line represents precision rate at different threshold settings

This graph makes it easy to see the relationship between precision and recall for any given threshold. The higher the threshold, the higher the precision, but the lower the recall. Note that high precision is required when you use IBM® Content Classification to generate automatic responses (these should be as accurate as possible).

The ideal threshold setting is the highest possible recall and precision rate. This goal is not always achievable, because the higher the recall rate, the lower the precision rate, and vice versa. Setting the most appropriate threshold for a category is a trade-off between these two rates. If there is a high precision rate with stringent scoring requirements, the Content Classification will miss many items that belong to the category. In order to be precise, the system will not include many items in the category that should be included. However, if there is a high recall rate with many items being received, there will be less accuracy.