We are facing an issue in ADM model learning. We have an existing model learning with channel and direction context and location as predictor. Recently I have created a new ADM model instance with Location, Channel and direction as context and no location predictor. We want to use propensity value from old model and new model should only learn from customer interactions. After adding these changes, when we get a response back for an interaction, old model is learning twice and new model is learning correctly and working fine. I have reverted back my changes and old model is learning correctly. Any idea, why is this behaviour for location as context. PFA of screenshots taken for this issue.
"ADM Digital" model should have executed twice in your Strategy execution logic. You may need to fix the Strategy execution logic.
you can refer to the "pxModelsExecuted" property details to understand the list of ADM Models executed for each Strategy Result (list of ADM model executed will be separated by ','). Please make sure the each ADM executed only once in your execution path to address the duplicate ADM learning(s).
Nanjundan Chinnasamy | PEGA Lead Decisioning Architect
Agree on Nanjundan's reply. In addition, if you really don't want the first model to pick up responses (perhaps because you want to use them in a sort of "read only" mode), you can always change the outcome labels for that model so they don't match with the actual responses.