Missing values in predictor
Hi All,
We have noticed that few of our predictors have missing values. What percentage of missing values is acceptable for a predictor? Do we have any threshold?
Regards,
Nizam
***Edited by Moderator Marije to add Capability tags***
Pegasystems Inc.
NL
Hi Nizam,
There's no limit here, but generally the predictive power of a predictor will be negatively influenced when it has more missing values and the adaptive models will in that case probably have other predictors become active instead.
I would keep an eye out for an increase of missing values over time for a predictor, which could point to an issue with the data feeding in. BTW note that in 8.7 there are new capabilities to monitor model predictors, see https://docs-previous.pega.com/platform-release-notes/87/whats-new-decision-management-87
Regards,
Ivar
Tech Mahindra
IN
@Ivar_Siccama Can we create this report from predictor snapshot table in lower versions of Pega? We haven't migrated to 8.7.