Some questions about data preprocessing

[Tonywhitemin]

Hi @pplonski,
Wish you have a good day!
Since I found that there are some good functions in mljar-AutoML for data preprocessing , but I don't know how to use it...
Could you help?

1. LeaveOneOutEncoder: The default setting is "categorical_to_int", May I know in what case will LeaveOneOutEncoder work?
   

2. Fill missing values with mean value:The default setting is "FILL_NA_MEDIAN". Can we specify "FILL_NA_MEAN" or others instead?
   

3. Scale method for "scale_log_and_normal" : The default setting is "scale_normal". Can we specify "scale_log_and_normal" instead? Or in what case will "scale_log_and_normal" work?
   

Thanks for your time!

[pplonski]

1. LeaveOneOutEncoder is not used - it was causing target leakage and overfitting.
2. FILL_NA_MEAN is not used - there is no big difference in results. You can try to overwrite default settings in your fork.
3. scale_log_and_normal is used when the target is skewed (there is a skew test done on the target).

[Tonywhitemin]

Thank you @pplonski !