Question
Areteans Technology Solutions Australia Pty Ltd.
IN
Last activity: 6 Sep 2021 7:17 EDT
Add existing Predictor in the Predictive model from Prediction Studio
Hi Team,
While creating a predictive model from historical dataset, through Predictive Studio of Pega , I have analysed that Pega is by default suggesting Predictor group(Group 1 , 2 etc) and the predictor list under each Group followed by two default Models ( Regression and Random Forest).
While grouping the Predictor , I have seen the default process is excluding few Predictors from the model, which has some business value or statistically correlated with Target.
My question is how can we append those predictors(this is not derived predictor , existing predictor) with the existing list for each Model?
I am using CDH using Pega v 8.4.
Pegasystems Inc.
IN
In the Pega wizard to build the model, you should be able to select additional predictors (you can select more than one predictor from every group). You need to make sure that the setting selected is Use all predictors
Areteans Technology Solutions Australia Pty Ltd.
IN
@SHRIDHAR Thanks for the responses. I am creating a Predictive model trained by csv data as source with 10+ original predictors, before finalize the Algorithm ( Regression or Random Forest etc ) it ask for selected the predictor from list of groups which you mentioned , but the group contains only 7-8 predictors excluding the other 3-4 predictors. But i need to include the other 3-4 predictors within the model. What configuration do I need to implement.
Pegasystems Inc.
IN
The predictors may have been excluded/disabled at the data analysis stage (they are excluded if the predictor performance is less than a threshold). You can reduce the threshold performance at the data analysis stage and these predictors can then be selected for the next step
Bear in mind that, though you can add these fields as predictors, they can end up reducing the performance of the model in the test set of the project. These fields are deselected because the system thinks they do not value to the model and adding them can reduce the performance of the model