Calculate classification accuracy
Calculate the classification accuracy for your solution and view the results in an interactive result matrix window.
The classification accuracy calculation compares the actual document types (ground truth) against the predicted classifications to provide detailed metrics, including F1 Score, Precision, and Recall. The result matrix helps you identify misclassification patterns and improve your solution's accuracy.
Before you can calculate classification accuracy, ensure that the following steps are complete:
-
Upload test documents and mark them for classification accuracy testing
-
Save your solution
-
On the
Documents page, select View classification
accuracy in the header.
The Classification Accuracy Details window is displayed.
-
Select Calculate Classification Accuracy.
When prompted to confirm the classification accuracy calculation with a time estimate, select Continue. To return to the Classification Accuracy Details window, select Cancel.
A progress indicator is displayed.
Optional. Select Cancel to stop the calculation if it is taking longer than expected.
When the classification accuracy calculation is complete, the Classification Accuracy Details window is updated with the new results, including:
-
Overall accuracy metrics (F1 Score, Precision, Recall)
-
Document counts (total tested, correctly classified, misclassified)
-
Timestamp of the calculation
-
Interactive result matrix showing actual versus predicted classifications
-
-
Review the classification accuracy results.
For detailed information about interpreting the metrics and confusion matrix, see Classification accuracy results.
- Optional. Select any cell with a number to navigate directly to the document type, which enables you to quickly adjust the training documents as needed.
-
If the accuracy is lower than expected, improve classification by adding more training documents for document types with high misclassification rates.
-
Review the result matrix to identify which document types are being confused with each other, and adjust your classification settings or training documents accordingly.
-
Adjust your classification settings in and recalculate the accuracy to measure improvement.