Adaptive Decision Manager Settings
I am working with adaptive models in Pega and have a few queries which are as below:
1. There will be several adaptive models getting generated as a result of a strategy run. Where do they get generated? Do they take up space in Pega DB or Cassandra?
2. When adaptive models get refreshed or when data analysis happens (as mentioned in Settings tab in adaptive model), does it have any adverse effect on any real time interaction?
3. Is ADM a system remote to Pega, leveraging cloud?
4. How does the refresh of adaptive models happen according to the settings? e.g. if the configuration is like : run data analysis after 500 and refresh after 150 interactions,
how does this get triggered when the count is reached?
5. Does ADM use Pega interaction History for creating patterns and evaluating pyPropensity?
5. The pxAdaptiveAnalytics which is a view on ADM, is it pointing to Pega DB or Cassandra?
It will be extremely helpful if I get clarified regarding these. Thanks in advance.
***Edited by Moderator: Pallavi to update platform capability tags***
Pegasystems Inc.
NL
Hi ANIKETD5,
Please find my answers below,
BTW which platform version are you using or will you be using?
Regards,
Ivar
Hi ANIKETD5,
Please find my answers below,
BTW which platform version are you using or will you be using?
Regards,
Ivar
1. There will be several adaptive models getting generated as a result of a strategy run. Where do they get generated? Do they take up space in Pega DB or Cassandra?
Both. The database tables are described in:
https://community.pega.com/knowledgebase/articles/adaptive-decision-manager-data-model
And Cassandra is used to keep a cache of the models.
2. When adaptive models get refreshed or when data analysis happens (as mentioned in Settings tab in adaptive model), does it have any adverse effect on any real time interaction?
No, the updating of the models is decoupled from the scoring and is usually configured to run on reserved nodes. BTW from version 7.3. there’s only one setting to trigger updates.
3. Is ADM a system remote to Pega, leveraging cloud?
ADM executes on Pega node(s). From version 7.3 there can be multiple ADM worker nodes providing resiliency and scalability.
4. How does the refresh of adaptive models happen according to the settings? e.g. if the configuration is like : run data analysis after 500 and refresh after 150 interactions,
how does this get triggered when the count is reached?
A data analysis or model refresh will be done after another 500 or 150 new responses respectively have been received. As noted above, from version 7.3 there’s only 1 setting that always does both and is called a model update.
5. Does ADM use Pega interaction History for creating patterns and evaluating pyPropensity?
You can always add your own (parameterized) predictors and these can be derived from IH data.
From version 8.1 we’ve introduced a very easy and efficient way of using IH data: https://community.pega.com/knowledgebase/articles/add-predictors-based-interaction-history
5. The pxAdaptiveAnalytics which is a view on ADM, is it pointing to Pega DB or Cassandra?
The pxAdaptiveAnalytics data set is used to handle responses. Responses are kept for a limited time in Cassandra.
Cognizant Technology Solutions
IN
Hi Ivar,
Thanks a lot for responding . They are very useful .
I am using Pega Marketing 7.22 on PRPC V7.2.2.
One more question: What exact differences should I expect when I change the settings of Adaptive Model i.e. let's say, if the "Run Data Analysis after" and "Refresh After" is 2 response, what kind of changes in system behaviour should I be expecting?
Thanks,Aniket.