Hello,
When using SimpleImputer with parameters SimpleImputer(handle_unknown=“use_encoded_value”, unknown_value=-1), I always get an error of the form :
init() got an unexpected keyword argument ‘handle_unknown’
If I omit the parameters with SimpleImputer(), I get the following form of error (wich is logical):
(Found unknown categories … in column … during transform).
How can I solve that?
Hi alidaher,
SimpleImputer() has no parameter handle_unknown.
OrdinalEncoder() has this parameter.
imputer = SimpleImputer(strategy="constant", fill_value="missing")
encoder = OrdinalEncoder(handle_unknown="use_encoded_value", unknown_value=-1)
categorical_processor = make_pipeline(imputer, encoder)
https://scikit-learn.org/stable/modules/generated/sklearn.impute.SimpleImputer.html
Hi,
Sorry, yes it’s the OrdinalEncoder that results in this error in all my codes.
When using OrdinalEncoder with parameters handle_unknown=“use_encoded_value”, unknown_value=-1), I always get an error when executing the encoder code, with the following form :
TypeError: init() got an unexpected keyword argument 'handle_unknown’
If I omit the parameters, I get the following form of error when executiong the cross-val on the testing data ((wich is logical because there are some new unknown values in the testing data)
ValueError: Found unknown categories [’ Holand-Netherlands’] in column 7 during transform
Thank you for your help.
When I use OrdinalEncoder().get_params() , I get only two parameters:
{‘categories’: ‘auto’, ‘dtype’: numpy.float64}
I am not getting the two other parameters:
{handle_unknown and unknown_value}
It seems that you are using an older version of scikit-learn. Are you running this code in a cell of the jupyter notebooks of the Fun MOOC platform or on an alternative jupyter installation?
To know which scikit-learn version you are running, you can type in a cell:
import sklearn
print(sklearn.__version__)
On the Fun MOOC platform, the currently installed version of scikit-learn is: 0.24.2.
I am using Jupyter Notebook ANACONDA. How can we update the sklearn version?
No problem with jupyter notebooks of the Fun MOOC.
If I try to update ANACONDA,SCIKIT-learn will be updated?
Thank you.
You can install scikit-learn from the conda-forge channel which is up to date:
https://anaconda.org/conda-forge/scikit-learn
On the command line this is:
conda install -c conda-forge scikit-learn
If you prefer to use the Anaconda Navigator, you can follow those instructions:
If you plan to do so, it’s probably a good idea to use a dedicated conda environment to install conda-forge packages to avoid mixing them with the packages of the main channel:
https://docs.anaconda.com/anaconda/navigator/tutorials/manage-environments/
You can also find more details on how to setup a valid environment for the MOOC here:
Actually I checked and scikit-learn 0.24.2 is also available from in the main Anaconda channel:
Ok. Thank you very much
Yes it’s working now. Thank you.