Prediction strategies in Pega Customer Decision Hub version 8.4.x use pyCampaignOptimization flags to enable the response timeout mechanism for delayed learning of adaptive models in outbound runs. In later versions of Pega Customer Decision Hub, this flag is deprecated. You enable the response timeout by using a prediction-level setting.
Upgrading customized prediction strategies (in which the pyCampaignOptimization flag is enabled) from version 8.4.x to 8.5.x results in an error in which the pxHandleResponses data flow run does not process responses for these strategies.
- This issue affects you if you are on Pega Customer Decision Hub version 8.4.x, you use the Next-Best-Action Designer framework and you have customized prediction strategies, such as Predict Action Propensity and Predict Treatment Propensity.
- Delayed learning does not work for you after you upgrade your application to version 8.5.x.
To prevent this issue, remove the pyCampaignOptimization flag and other associated properties from your prediction strategies before you upgrade your application:
Identify the strategies that contain Set Property shapes which contain the pyCampaignOptimization flag.
In Prediction Studio, open a prediction, and then click.
On the strategy canvas, double-click each Set Property shape, and then check whether the configuration contains any of the following properties:
In the prediction strategies that you identified, remove the following properties:
To use the response timeout mechanism for outbound runs, note the strategies that you updated and the associated predictions. After you upgrade your application, set the response timeout in the prediction settings.For more information, see Setting the response timeout.