121 lines
3.9 KiB
Python
121 lines
3.9 KiB
Python
|
# Sebastian Raschka 2014-2020
|
||
|
|
# mlxtend Machine Learning Library Extensions
|
||
|
|
#
|
||
|
|
# Object for selecting a dataset column in scikit-learn pipelines.
|
||
|
|
# Author: Sebastian Raschka <sebastianraschka.com>
|
||
|
|
#
|
||
|
|
# License: BSD 3 clause
|
||
|
|
|
||
|
|
|
||
|
|
from sklearn.base import BaseEstimator
|
||
|
|
import numpy as np
|
||
|
|
|
||
|
|
|
||
|
|
class ColumnSelector(BaseEstimator):
|
||
|
|
"""Object for selecting specific columns from a data set.
|
||
|
|
|
||
|
|
Parameters
|
||
|
|
----------
|
||
|
|
cols : array-like (default: None)
|
||
|
|
A list specifying the feature indices to be selected. For example,
|
||
|
|
[1, 4, 5] to select the 2nd, 5th, and 6th feature columns, and
|
||
|
|
['A','C','D'] to select the name of feature columns A, C and D.
|
||
|
|
If None, returns all columns in the array.
|
||
|
|
|
||
|
|
drop_axis : bool (default=False)
|
||
|
|
Drops last axis if True and the only one column is selected. This
|
||
|
|
is useful, e.g., when the ColumnSelector is used for selecting
|
||
|
|
only one column and the resulting array should be fed to e.g.,
|
||
|
|
a scikit-learn column selector. E.g., instead of returning an
|
||
|
|
array with shape (n_samples, 1), drop_axis=True will return an
|
||
|
|
aray with shape (n_samples,).
|
||
|
|
|
||
|
|
Examples
|
||
|
|
-----------
|
||
|
|
For usage examples, please see
|
||
|
|
http://rasbt.github.io/mlxtend/user_guide/feature_selection/ColumnSelector/
|
||
|
|
|
||
|
|
"""
|
||
|
|
|
||
|
|
def __init__(self, cols=None, drop_axis=False):
|
||
|
|
self.cols = cols
|
||
|
|
self.drop_axis = drop_axis
|
||
|
|
|
||
|
|
def fit_transform(self, X, y=None):
|
||
|
|
""" Return a slice of the input array.
|
||
|
|
|
||
|
|
Parameters
|
||
|
|
----------
|
||
|
|
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
|
||
|
|
Training vectors, where n_samples is the number of samples and
|
||
|
|
n_features is the number of features.
|
||
|
|
y : array-like, shape = [n_samples] (default: None)
|
||
|
|
|
||
|
|
Returns
|
||
|
|
---------
|
||
|
|
X_slice : shape = [n_samples, k_features]
|
||
|
|
Subset of the feature space where k_features <= n_features
|
||
|
|
|
||
|
|
"""
|
||
|
|
return self.transform(X=X, y=y)
|
||
|
|
|
||
|
|
def transform(self, X, y=None):
|
||
|
|
""" Return a slice of the input array.
|
||
|
|
|
||
|
|
Parameters
|
||
|
|
----------
|
||
|
|
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
|
||
|
|
Training vectors, where n_samples is the number of samples and
|
||
|
|
n_features is the number of features.
|
||
|
|
y : array-like, shape = [n_samples] (default: None)
|
||
|
|
|
||
|
|
Returns
|
||
|
|
---------
|
||
|
|
X_slice : shape = [n_samples, k_features]
|
||
|
|
Subset of the feature space where k_features <= n_features
|
||
|
|
|
||
|
|
"""
|
||
|
|
|
||
|
|
# We use the loc or iloc accessor if the input is a pandas dataframe
|
||
|
|
if hasattr(X, 'loc') or hasattr(X, 'iloc'):
|
||
|
|
if type(self.cols) == tuple:
|
||
|
|
self.cols = list(self.cols)
|
||
|
|
types = {type(i) for i in self.cols}
|
||
|
|
if len(types) > 1:
|
||
|
|
raise ValueError(
|
||
|
|
'Elements in `cols` should be all of the same data type.'
|
||
|
|
)
|
||
|
|
if isinstance(self.cols[0], int):
|
||
|
|
t = X.iloc[:, self.cols].values
|
||
|
|
elif isinstance(self.cols[0], str):
|
||
|
|
t = X.loc[:, self.cols].values
|
||
|
|
else:
|
||
|
|
raise ValueError(
|
||
|
|
'Elements in `cols` should be either `int` or `str`.'
|
||
|
|
)
|
||
|
|
else:
|
||
|
|
t = X[:, self.cols]
|
||
|
|
|
||
|
|
if t.shape[-1] == 1 and self.drop_axis:
|
||
|
|
t = t.reshape(-1)
|
||
|
|
if len(t.shape) == 1 and not self.drop_axis:
|
||
|
|
t = t[:, np.newaxis]
|
||
|
|
return t
|
||
|
|
|
||
|
|
def fit(self, X, y=None):
|
||
|
|
""" Mock method. Does nothing.
|
||
|
|
|
||
|
|
Parameters
|
||
|
|
----------
|
||
|
|
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
|
||
|
|
Training vectors, where n_samples is the number of samples and
|
||
|
|
n_features is the number of features.
|
||
|
|
y : array-like, shape = [n_samples] (default: None)
|
||
|
|
|
||
|
|
Returns
|
||
|
|
---------
|
||
|
|
self
|
||
|
|
|
||
|
|
"""
|
||
|
|
return self
|