Can I add new features in an existing dataset using function transformers in scikit-learn

Question

I have written a code that can add 3 new columns into a NumPy array, using function transformer(1 st column is element-wise +, 2nd is element-wise *, 3rd is element-wise /. Just need to know if in this way I can add new features to an existing dataset

import numpy as np
from sklearn.preprocessing import FunctionTransformer
def col_add(x):
    x1 = x[:, 0] + x[:, 1]
    x2 = x[:, 0] * x[:, 1]
    x3 = x[:, 0] / x[:, 1]
    return np.c_[x, x1, x2, x3]
col_adder = FunctionTransformer(col_add)
arr = np.array([[2, 7], [4, 9], [3, 5]])
arr
array([[2, 7],
       [4, 9],
       [3, 5]])
col_adder.transform(arr) # will add 3 columns
array([[ 2.        ,  7.        ,  9.        , 14.        ,  0.28571429],
       [ 4.        ,  9.        , 13.        , 36.        ,  0.44444444],
       [ 3.        ,  5.        ,  8.        , 15.        ,  0.6       ]])

You asked if this can add new features, then provided code that shows that indeed it has added new features. What exactly is the question? — Ben Reiniger, Dec 16 '21 at 16:37

score 1 · Answer 1 · answered Jun 22 '21 at 02:57

Scikit-learn FeatureUnion is what you are looking for:

import numpy as np
from sklearn.preprocessing import FunctionTransformer
from sklearn.pipeline import FeatureUnion
from sklearn.datasets import load_iris
X, _ = load_iris(return_X_y = True)
def col_add(x):
    x1 = x[:, 0] + x[:, 1]
    x2 = x[:, 0] * x[:, 1]
    x3 = x[:, 0] / x[:, 1]
    return np.c_[x, x1, x2, x3]
transformer = FunctionTransformer(col_add)
f = FeatureUnion([("t",transformer)]).fit(X)
Xnew = f.transform(X)

Can I add new features in an existing dataset using function transformers in scikit-learn

1 Answers1