Skip to main content
def Predictor.fe_stabilize(self, data, preprocess=False) [source]
Apply trained time-stability filtering to input data. Removes variables that were detected as unstable (time-variant distributions) during training. Ensures only temporally stable features propagate downstream.

​
Parameters

data : pDF
Input data of shape (n_samples, n_features).
 preprocess : bool
If True, accepts original columns and applies
fe_clean first. If False, expects output
from fe_clean.

​
Returns

FeStabilizerResult