Reviewing Predictive Model Mis-predictions and Providing Feedback
In 7.4 how can I easily see when a model mis-predicted and flag it as something that should be added back to the training data?
I was looking into how to make this in the channel and interfaces tab, but what if we want to display this to a stakeholder SME and ask them to review the model mis-predictions and notify us which ones were truly misses and needed to be added as training data to the model.
Since the content would be text its easy to potentially add bad data to the training data if I just take every misprediction at face value without getting an SME's opinion.
Pegasystems Inc.
US
Hello Garrett,
SME review is possible with the Channels and Interfaces screen in the training data tab.
Here all the incoming records are stored where the SME can review before adding the learnings back to the model as feedback.
Regards,
Shantanu