-
Notifications
You must be signed in to change notification settings - Fork 35
Open
Description
Hi,
I have been running the tutorial.
and I have approximately the same average score after tpot than you (more around 0.79 but close).
I was wondering how you got this information:
https://github.com/chriotte/Stress_classifier_with_AutoML_and_wearable_devices/blob/master/images/dataoverview.png
is that tpot who "told you" that avnn was rejected because highly correlated with interval in seconds etc ... ?
should not it be then trained again without avnn, for training it should not have two highly correlated features normally.
then if I want to run it on other dataset I don't have to calculate the avnn features ?
Best regards
PS: I don't find the ground truth for the viking dataset, is that normal ?
Metadata
Metadata
Assignees
Labels
No labels