Posted: 16 Jan 2018 9:14 EST Last activity: 27 Apr 2018 6:16 EDT
How to identify and remove badly performing predictor in DSM 7.31?
I got 30 predictors in my Adaptive Model. After few runs,I realized that some of the predictors are not performing well. Is it possible to identify those bad predictors in Pega DSM 7.31? Your help is very much appreciated.
Please note that the total list of 30 predictors, is a list of candidate predictors. ADM will decide for itself for each model (proposition) which predictors will become active. For more info, there's a Predictors 101 at: https://mesh.pega.com/docs/DOC-127801
Be careful to withdraw a predictor too quickly because for a proposition you might add in the future it might still have predictive power. I would only withdraw it from the list of candidate predictors if you're sure it's never going to be used. It doesn't do harm to have a larger set of candidate predictors (e.g. 200), although there is a slight performance overhead.