跳到内容

特征转换器

本文件是 TPOT 库的一部分。

当前版本的 TPOT 由 Cedars-Sinai 的以下人员开发：- Pedro Henrique Ribeiro (https://github.com/perib, https://www.linkedin.com/in/pedro-ribeiro/) - Anil Saini (anil.saini@cshs.org) - Jose Hernandez (jgh9094@gmail.com) - Jay Moran (jay.moran@cshs.org) - Nicholas Matsumoto (nicholas.matsumoto@cshs.org) - Hyunjun Choi (hyunjun.choi@cshs.org) - Gabriel Ketron (gabriel.ketron@cshs.org) - Miguel E. Hernandez (miguel.e.hernandez@cshs.org) - Jason Moore (moorejh28@gmail.com)

TPOT 的原始版本主要由宾夕法尼亚大学的以下人员开发：- Randal S. Olson (rso@randalolson.com) - Weixuan Fu (weixuanf@upenn.edu) - Daniel Angell (dpa34@drexel.edu) - Jason Moore (moorejh28@gmail.com) - 以及许多慷慨的开源贡献者

TPOT 是免费软件：您可以根据自由软件基金会发布的 GNU 宽通用公共许可证（第三版或您选择的任何更高版本）的条款重新分发和/或修改它。

TPOT 的分发是希望它有用，但不提供任何担保；甚至不提供适销性或特定用途适用性的默示担保。详情请参阅 GNU 宽通用公共许可证。

您应该已经随 TPOT 收到了一份 GNU 宽通用公共许可证的副本。如果没有，请参阅 https://gnu.ac.cn/licenses/。

CategoricalSelector

基础类：BaseEstimator, TransformerMixin

用于选择分类特征并使用 OneHotEncoder 对其进行转换的元转换器。

参数

名称 类型 描述 默认值
threshold int

每个特征的最大唯一值数量，用于判断特征是否为分类特征。

 10
minimum_fraction

特征中唯一值的最小比例，用于判断特征是否为分类特征。

 None
源代码位于 tpot/builtin_modules/feature_transformers.py 中 
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
class CategoricalSelector(BaseEstimator, TransformerMixin):
    """Meta-transformer for selecting categorical features and transform them using OneHotEncoder.

    Parameters
    ----------

    threshold : int, default=10
        Maximum number of unique values per feature to consider the feature
        to be categorical.

    minimum_fraction: float, default=None
        Minimum fraction of unique values in a feature to consider the feature
        to be categorical.
    """

    def __init__(self, threshold=10, minimum_fraction=None):
        """Create a CategoricalSelector object."""
        self.threshold = threshold
        self.minimum_fraction = minimum_fraction


    def fit(self, X, y=None):
        """Do nothing and return the estimator unchanged
        This method is just there to implement the usual API and hence
        work in pipelines.
        Parameters
        ----------
        X : array-like
        """
        X = check_array(X, accept_sparse='csr')
        return self


    def transform(self, X):
        """Select categorical features and transform them using OneHotEncoder.

        Parameters
        ----------
        X: numpy ndarray, {n_samples, n_components}
            New data, where n_samples is the number of samples and n_components is the number of components.

        Returns
        -------
        array-like, {n_samples, n_components}
        """
        selected = auto_select_categorical_features(X, threshold=self.threshold)
        X_sel, _, n_selected, _ = _X_selected(X, selected)

        if n_selected == 0:
            # No features selected.
            raise ValueError('No categorical feature was found!')
        else:
            ohe = OneHotEncoder(categorical_features='all', sparse=False, minimum_fraction=self.minimum_fraction)
            return ohe.fit_transform(X_sel)

__init__(threshold=10, minimum_fraction=None)

创建一个 CategoricalSelector 对象。

源代码位于 tpot/builtin_modules/feature_transformers.py 中 
63
64
65
66
def __init__(self, threshold=10, minimum_fraction=None):
    """Create a CategoricalSelector object."""
    self.threshold = threshold
    self.minimum_fraction = minimum_fraction

fit(X, y=None)

不执行任何操作并返回未更改的估计器。此方法仅用于实现常用的 API，因此可在 pipeline 中工作。

参数

名称 类型 描述 默认值
X array - like
必需
源代码位于 tpot/builtin_modules/feature_transformers.py 中 
69
70
71
72
73
74
75
76
77
78
def fit(self, X, y=None):
    """Do nothing and return the estimator unchanged
    This method is just there to implement the usual API and hence
    work in pipelines.
    Parameters
    ----------
    X : array-like
    """
    X = check_array(X, accept_sparse='csr')
    return self

transform(X)

选择分类特征并使用 OneHotEncoder 对其进行转换。

参数

名称 类型 描述 默认值
X

新数据，其中 n_samples 是样本数量，n_components 是组件数量。

 必需

返回值

类型 描述
(array - like, {n_samples, n_components})
源代码位于 tpot/builtin_modules/feature_transformers.py 中 
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
def transform(self, X):
    """Select categorical features and transform them using OneHotEncoder.

    Parameters
    ----------
    X: numpy ndarray, {n_samples, n_components}
        New data, where n_samples is the number of samples and n_components is the number of components.

    Returns
    -------
    array-like, {n_samples, n_components}
    """
    selected = auto_select_categorical_features(X, threshold=self.threshold)
    X_sel, _, n_selected, _ = _X_selected(X, selected)

    if n_selected == 0:
        # No features selected.
        raise ValueError('No categorical feature was found!')
    else:
        ohe = OneHotEncoder(categorical_features='all', sparse=False, minimum_fraction=self.minimum_fraction)
        return ohe.fit_transform(X_sel)

ContinuousSelector

基础类：BaseEstimator, TransformerMixin

用于选择连续特征并使用 PCA 对其进行转换的元转换器。

参数

名称 类型 描述 默认值
threshold int

每个特征的最大唯一值数量，用于判断特征是否为分类特征。

 10
svd_solver string {'auto', 'full', 'arpack', 'randomized'}

auto：求解器根据 X.shape 和 n_components 的默认策略选择：如果输入数据大于 500x500 且要提取的组件数量小于数据最小维度的 80%，则启用更高效的 'randomized' 方法。否则，计算精确的完整 SVD，并在之后可选地截断。full：通过 scipy.linalg.svd 调用标准 LAPACK 求解器运行精确的完整 SVD，并通过后处理选择组件。arpack：通过 scipy.sparse.linalg.svds 调用 ARPACK 求解器运行截断到 n_components 的 SVD。它严格要求 0 < n_components < X.shape[1]。randomized：通过 Halko 等人的方法运行随机 SVD。

 'randomized'
iterated_power int >= 0，或 'auto'，(默认 'auto')

对于 svd_solver == 'randomized' 计算的幂方法的迭代次数。

 'auto'
源代码位于 tpot/builtin_modules/feature_transformers.py 中 
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
class ContinuousSelector(BaseEstimator, TransformerMixin):
    """Meta-transformer for selecting continuous features and transform them using PCA.

    Parameters
    ----------

    threshold : int, default=10
        Maximum number of unique values per feature to consider the feature
        to be categorical.

    svd_solver : string {'auto', 'full', 'arpack', 'randomized'}
        auto :
            the solver is selected by a default policy based on `X.shape` and
            `n_components`: if the input data is larger than 500x500 and the
            number of components to extract is lower than 80% of the smallest
            dimension of the data, then the more efficient 'randomized'
            method is enabled. Otherwise the exact full SVD is computed and
            optionally truncated afterwards.
        full :
            run exact full SVD calling the standard LAPACK solver via
            `scipy.linalg.svd` and select the components by postprocessing
        arpack :
            run SVD truncated to n_components calling ARPACK solver via
            `scipy.sparse.linalg.svds`. It requires strictly
            0 < n_components < X.shape[1]
        randomized :
            run randomized SVD by the method of Halko et al.

    iterated_power : int >= 0, or 'auto', (default 'auto')
        Number of iterations for the power method computed by
        svd_solver == 'randomized'.

    """

    def __init__(self, threshold=10, svd_solver='randomized' ,iterated_power='auto', random_state=42):
        """Create a ContinuousSelector object."""
        self.threshold = threshold
        self.svd_solver = svd_solver
        self.iterated_power = iterated_power
        self.random_state = random_state


    def fit(self, X, y=None):
        """Do nothing and return the estimator unchanged
        This method is just there to implement the usual API and hence
        work in pipelines.
        Parameters
        ----------
        X : array-like
        """
        X = check_array(X)
        return self


    def transform(self, X):
        """Select continuous features and transform them using PCA.

        Parameters
        ----------
        X: numpy ndarray, {n_samples, n_components}
            New data, where n_samples is the number of samples and n_components is the number of components.

        Returns
        -------
        array-like, {n_samples, n_components}
        """
        selected = auto_select_categorical_features(X, threshold=self.threshold)
        _, X_sel, n_selected, _ = _X_selected(X, selected)

        if n_selected == 0:
            # No features selected.
            raise ValueError('No continuous feature was found!')
        else:
            pca = PCA(svd_solver=self.svd_solver, iterated_power=self.iterated_power, random_state=self.random_state)
            return pca.fit_transform(X_sel)

__init__(threshold=10, svd_solver='randomized', iterated_power='auto', random_state=42)

创建一个 ContinuousSelector 对象。

源代码位于 tpot/builtin_modules/feature_transformers.py 中 
138
139
140
141
142
143
def __init__(self, threshold=10, svd_solver='randomized' ,iterated_power='auto', random_state=42):
    """Create a ContinuousSelector object."""
    self.threshold = threshold
    self.svd_solver = svd_solver
    self.iterated_power = iterated_power
    self.random_state = random_state

fit(X, y=None)

不执行任何操作并返回未更改的估计器。此方法仅用于实现常用的 API，因此可在 pipeline 中工作。

参数

名称 类型 描述 默认值
X array - like
必需
源代码位于 tpot/builtin_modules/feature_transformers.py 中 
146
147
148
149
150
151
152
153
154
155
def fit(self, X, y=None):
    """Do nothing and return the estimator unchanged
    This method is just there to implement the usual API and hence
    work in pipelines.
    Parameters
    ----------
    X : array-like
    """
    X = check_array(X)
    return self

transform(X)

选择连续特征并使用 PCA 对其进行转换。

参数

名称 类型 描述 默认值
X

新数据，其中 n_samples 是样本数量，n_components 是组件数量。

 必需

返回值

类型 描述
(array - like, {n_samples, n_components})
源代码位于 tpot/builtin_modules/feature_transformers.py 中 
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
def transform(self, X):
    """Select continuous features and transform them using PCA.

    Parameters
    ----------
    X: numpy ndarray, {n_samples, n_components}
        New data, where n_samples is the number of samples and n_components is the number of components.

    Returns
    -------
    array-like, {n_samples, n_components}
    """
    selected = auto_select_categorical_features(X, threshold=self.threshold)
    _, X_sel, n_selected, _ = _X_selected(X, selected)

    if n_selected == 0:
        # No features selected.
        raise ValueError('No continuous feature was found!')
    else:
        pca = PCA(svd_solver=self.svd_solver, iterated_power=self.iterated_power, random_state=self.random_state)
        return pca.fit_transform(X_sel)