Analyzing and exporting text analysis models

This is the fourth and final step of the Create text analysis models wizard. In this step, you can analyze the overall accuracy of each model, based on its F-score. You can also download the .XLSX file containing the model accuracy analysis and inspect the results in detail. After you review the build results, you can save each model to a .JAR extension file and export it. Optionally, you can also create a new text analysis model binary rule that includes the .JAR file with the selected model.

1. In the Analyze created models section, view the aggregated accuracy score of each model.
   The text analysis model accuracy is based on the F-score.
   Note: The F-score is the most commonly used measure in the field of text analysis and is the harmonic average of the precision and recall values of the analyzed sample. The F-score can have values ranging from 0 to 1 (1 represents the highest model accuracy).
2. Optional: Review the detailed results of the model accuracy analysis.

   1. In the Download model analysis area, click the link to download the .XLSX file that contains the detailed model accuracy analysis for each model that you created.
      Note: The name of the model accuracy analysis file matches the name of the training data file.
   2. Save the file in your directory.
   3. Open the file and review the detailed model accuracy analysis. The .XLSX file has the following structure:
      • DATA_SHEET
      • Classification_<ALGORITHM_NAME>
      • SCORE_SHEET_<ALGORITHM_NAME>
3. Optional: Export the model.

   1. In the Model to export area, expand the drop-down list and select a model.
   2. Click Save to file to download the .JAR file with the selected model to your local directory for further processing.
   3. Optional: Click Cancel to abort and close the wizard.
      Perform this step only when you do not want to create a text analysis model binary rule file. If you abort the wizard, no work object is created in the Machine Learning Modeling tab.
4. Optional: Create a text analysis model binary rule file in the Pega 7 Platform.

   1. Select Include model in new rule file.
   2. Specify the Label, Ruleset, and RuleSet version parameters of the new rule file.
   3. Click Create.

   The system creates a model binary rule file that contains the text analysis .JAR extension file. Once you create a model binary rule file, you can view it in the Machine Learning Modeling tab. You can also upload the model binary rule file as part of the following decision data:

   • Taxonomy (for the classification analysis models)
   • Sentiment Model (for the sentiment analysis models)

   Note: The model binary file can be exported to branches.

Open topic with navigation