Question
Accenture
US
Last activity: 9 Sep 2020 17:13 EDT
How can we set nuetral outcomes in Adaptive models?
Hi,
In my current scenario we have a couple of positive outcomes and a neutral one, but I do not have a truly negative outcome.
How can I set these values in Adaptive model without botching up the model learning?
Thanks,
Chetan
Pegasystems Inc.
IN
Can you please try below:
Accenture
US
Thanks for the reply, but the help section doesn't have the information I am looking for.
I am looking for 2 things:
1.How to avoid setting a negative outcome?
2.How to set a neutral outcome?
I am looking for these things because the responses I get are 2 positive and 1 neutral, no negative responses as of now.
Regards,
Chetan
Pegasystems Inc.
NL
Banner placement is a typical use case where there are no explicit negatives: there are just clicks but no "not clicks". There's many such use cases.
The recommended pattern is to post a "negative" when the banner is shown ("impressed"). Ideally only when it is viewable by the customer, but if that's impossible even just a "negative" to all the top N decisions is an okay proxy.
Pegasystems Inc.
US
In a slightly related use case, when an email contains multiple offers, should the email open, and the click-though on the individual offers both be positive outcomes for each offer - assuming the email delivery is a negative outcome for each offer?
Merkle Inc
US
1) In my opinion, the term 'NEGATIVE' is misleading, as the Adaptive models are not predicting a negative behavior. The right terms should be 'POSITIVE' and 'NON-POSITIVE'
2) So these are valid ways to assign the positives and 'negatives' on depending on the use cases:
- Web ad: Positive=Clickthough, Negative=Impression without clicks (make sure you don't double count the impression)
- Email ad: Positive = Clickthrough or Open (Pick One), Negative = Impression without clicks (or opens)
Are above assumptions correct?