Hands-On Data Analysis with Pandas

Additional transformers

What if, rather than scaling our data or encoding it, we want to run a mathematical operation, like taking the square root or the logarithm? The preprocessing module also has some classes for this. While there are a few that do a specific transformation, like the QuantileTransformer, we will focus our attention on the FunctionTransformer, which lets us provide an arbitrary function to use:

>>> from sklearn.preprocessing import FunctionTransformer>>> FunctionTransformer(...     np.abs, validate=True... ).fit_transform(X_train.dropna())array([[0.073 , 2.6   , 1.6   ],
       ...,
       [0.249 , 4.62  , 2.99  ]])

Here, we took the absolute value of every number. Notice the validate=True argument; the FunctionTransformer knows that scikit-learn ...