doesn't catch some propocessing error of skelarn

[caprone]

"ValueError: Minimum of desired feature range must be smaller than maximum. Got (0.0, 0.0)." after useless 3 hours of training on a very smaller dataset...too slow a, not verbose...and bugged

[Javihaus]

Can I see the dataset?

[filiu97]

I've had the same error. Could you post the fix for it?

Thanks.

[caprone]

same error...

[caprone]

ok, seems that issue is only for zero variance features...then MinMaxScaler raises that ValueError;
we need to delete all zero variance features from our dataset(that is ever a good idea):

def remove_constant_value_features(df):
return [e for e in df.columns if df[e].nunique() == 1]

feat to remove:

drop_col = remove_constant_value_features(train)