Scikit-learn 具有多个值的输入

Question

Is there a way for a Scikit-learn Imputer to look for and replace multiple values which are considered "missing values"?

For example, I would like to do something like

imp = Imputer(missing_values=(7,8,9))

But according to the docs, the missing_values parameter only accepts a single integer:

missing_values: integer or “NaN”, optional (default=”NaN”)

The placeholder for the missing values. All occurrences of missing_values will be imputed. For missing values encoded as np.nan, use the string value “NaN”.

Answer 1

Why not to do this manually in your original dataset? Assuming you are using pd.DataFrame you can do the following:

import numpy as np
import pandas as pd
from sklearn.preprocessing import Imputer

df = pd.DataFrame({'A': [1, 2, 3, 8], 'B': [1, 2, 5, 3]})
df_new = df.replace([1, 2], np.nan)
df_imp = Imputer().fit_transform(df_new)

This results in df_imp :

array([[ 5.5,  4. ],
   [ 5.5,  4. ],
   [ 3. ,  5. ],
   [ 8. ,  3. ]])

If you want to make this a part of a pipeline, you would just need to implement a custom transformer with a similar logic.

Answer 2

You could chain multiple imputers in a pipeline, but that might become hectic pretty soon and I'm not sure how efficient that is.

pipeline = make_pipeline(
    SimpleImputer(missing_values=7, strategy='constant', fill_value=10),
    SimpleImputer(missing_values=8, strategy='constant', fill_value=10),
    SimpleImputer(missing_values=9, strategy='constant', fill_value=10)
)