Skip to main content

Extract historical responses from adaptive models for offline analysis (8.4)

Suggest edit Updated on February 24, 2020

You can now analyze historical customer responses and learn more about adaptive models by extracting past interactions. Based on the historical responses that you extract and download, you can also build a model in a machine learning service of your choice.

Every interaction between your customer and your business, such as accepting or rejecting an offer or clicking your banner on a website, provides a meaningful context for your adaptive models. By enabling the recording of past interactions in Prediction Studio, you can reuse the gathered historical data to improve future adaptive models and, in effect, increase the accuracy of customer decisions. When enabling historical data recording, you can also customize the data that you collect by specifying the percentage of all positive and negative responses that you want to record.

The following video demonstrates how to enable historical data recording for an adaptive model that predicts whether a customer is likely to click a web banner. A web banner typically has a significantly lower number of positive responses (banner clicks), than negative responses (banner impressions). In such cases, you can decide to record all positive responses (100%) and only a small amount of negative responses (1%).

"The video demonstrates how to enable recording customer responses to a web banner by opening the adaptive model, and then, in the Settings tab, selecting the Record historical data check box. You then specify the positive outcome sample percentage to 100%, and the negative outcome sample percentage to 1%"
Enabling historical data recording

For more information, see Extracting historical responses from adaptive models.

Did you find this content helpful? YesNo

Have a question? Get answers now.

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

Ready to crush complexity?

Experience the benefits of Pega Community when you log in.

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