Posted: 15 May 2018 4:26 EDT Last activity: 15 May 2018 4:34 EDT
Tracking performance of external models through Adaptive
Adaptive Models can be used to track performance of external (PAD, PMML) models, Scorecards or even strategies.
In the graph below, you see how the performance of a PAD model is tracked
The whole configuration is straightforward: we are feeding the outputs of the off-line model (as .pxSegment) as the only predictor into an adaptive model via a parameter. The grouping threshold of the Adaptive Model has been set to a minimal value so to make sure all distinct values of .pxSegment will be treated seperately.
For the reporting part, we just store the performance (which is added to the default outputs of the model in the strategy shape configuration "Enable additional mapping") in a database table on which a standard Report Definiton is run.