Community Tip - You can change your system assigned username to something more personal in your community settings. X
Is there a way to find out what the best learner is after training a predictive model? I haven't been able to find it in the documentation or searching the menus/services.
Also, for the elite average ensemble, it's stated in the docs the it chooses the best n learners, but it's never mentioned how it decides what the n number is. There's mention that they're chosen based on their predictive score, but how does it determine the cutoff between good and bad learners?
It'l be helpful to know a little more about what the black box is doing in order to improve further models.
Solved! Go to Solution.
When you select the Enemble technique, you also specify the metric by which the learners are compared and selected, e.g. by RMSE.
You might also be interested in https://community.ptc.com/t5/IoT-Tips/ThingWorx-Analytics-Training-Module-6-Part-2/ta-p/843321 starting at 6:05
For the learners selected, the article https://www.ptc.com/en/support/article/CS258769 gives some information about it.
So from the exported pmml file you should be able to see what and how many learners have been used
Hope this helps
Christophe
Thanks for your answer Cristophe, I actually haven't seen that article and it answers some of my questions so thank you for that. However it still doesn't tell me how does the ensemble logic chooses the model to be used. I checked some exported pmml files from models that had more than one learner, but it still only chose 1 best and went with that, so it clearly has a cutoff score or something of that matter.
Any other material that can answer that?
When you select the Enemble technique, you also specify the metric by which the learners are compared and selected, e.g. by RMSE.
You might also be interested in https://community.ptc.com/t5/IoT-Tips/ThingWorx-Analytics-Training-Module-6-Part-2/ta-p/843321 starting at 6:05
Thank you Rocko, with your and Cristophe's answer I got what I needed. Regards,