Skip to main content


         This documentation site is for previous versions. Visit our new documentation site for current releases.      
 

Configuring the monitoring of model input and output

Updated on July 5, 2022

Monitor input, that is, predictors, and output of your models and predictions to observe whether they behave as you expect.

Models use predictors, such as age, income, location, or current product subscriptions, to predict customer responses or case outcomes. The output is the data that model generate, including propensities, numeric output for continuous models, and symbolic output (labels) for binary and categorical models.

Monitoring predictors and output can provide useful insights about the performance of your predictions and models. It helps in the early stages of model development, before you capture any responses. It also helps for certain model types that do not receive responses regularly or at all, such as predictive models.

Note: The monitoring of predictors and output is enabled by default, and the system is configured to monitor 5% of model executions. Monitoring uses the Stream service and may require an extension of the Stream service disk space.
Before you begin:

Review the information about the estimated payload for model monitoring, which determines the amount of model executions that you can monitor without a negative effect on system performance.

Configure the analytics repository:

  • Ensure that the repository has 15 GB of disk space.
  • Do not use the defaultstore temporary repository, but a more stable one.

Ensure that the disk space allocated to the Stream service is 250 GB.

  1. In the navigation pane of Prediction Studio, click SettingsPrediction Studio settings.
  2. In the Monitor model input and output section, ensure that the Monitor model input and output data checkbox is selected.
  3. In the Monitor percentage field, define the amount of data that you want to monitor.
    Note: The default monitor percentage is 5%, which means that the system monitors 5% of all model executions. Higher values might affect performance. Choose a value that meets your monitoring requirements and ensures good system performance. For more information, see Estimating the model monitoring payload.
  4. In the Monitoring coverage list, select which data you want to monitor:
    • Monitor output (propensities, numeric values, labels) and input (predictors) by selecting Monitor output and input.
    • Monitor output only (propensities, numeric values, labels) by selecting Monitor output only.
    You can find the metrics on the Analysis tab in predictions. For more information, see Monitoring predictions.
  5. Optional: To record and save the monitoring data in the analytics repository, select Record monitoring data.
    Monitor model input and output settings in Prediction Studio
    A Prediction Studio screen displaying settings that contain parameters for monitoring model input and output.
  6. Click Save.
  • Estimating the model monitoring payload

    Starting in Pega Platform version 8.7, the monitoring of model predictors (input) and output is enabled by default. The system requirements to support this feature are 250 GB of Stream service disk space and 15 GB of analytics repository disk space.

Have a question? Get answers now.

Visit the Support Center to ask questions, engage in discussions, share ideas, and help others.

Did you find this content helpful?

Want to help us improve this content?

We'd prefer it if you saw us at our best.

Pega.com is not optimized for Internet Explorer. For the optimal experience, please use:

Close Deprecation Notice
Contact us