数据挖掘特征提取方法-汇集

1.基于树模型提取特征

# 使用树模型提取特征
import numpy as np
from sklearn import feature_selection
from sklearn.ensemble import GradientBoostingClassifier

matrix = np.array(X)
target = np.array(target)
temp = feature_selection.SelectFromModel(GradientBoostingClassifier()).fit(matrix, target)
indx = temp._get_support_mask().tolist()
scores = get_importance(temp.estimator_).tolist()
result = temp.transform(matrix).tolist()
return scores, indx, result

# X: array-like
# target: array-like
# http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectFromModel.html

2.基于L1，L2惩罚值提取特征

# 基于L1,L2惩罚值提取特征
import numpy as np
from sklearn import feature_selection
from sklearn.linear_model import LogisticRegression

matrix = np.array(arr0)
target = np.array(target)
temp = feature_selection.SelectFromModel(LogisticRegression(penalty="l1", C=0.1)).fit(matrix, target)
indx = temp._get_support_mask().tolist()
scores = get_importance(temp.estimator_).tolist()
result = temp.transform(matrix).tolist()
return scores, indx, result

# X: array-like
# target: array-like
# http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectFromModel.html

3.递归特征消除法提取特征

# 递归特征消除法
import numpy as np
from sklearn import feature_selection
from sklearn.linear_model import LogisticRegression

matrix = np.array(X)
target = np.array(target)
temp = feature_selection.RFE(estimator=LogisticRegression(), n_features_to_select=n_features).fit(matrix, target)
scores = temp.ranking_.tolist()
indx = temp.support_.tolist()
result = temp.transform(matrix).tolist()
return scores, indx, result

# X: array-like
# target: array-like
# n-features: int
# http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html

4.互信息选择法提取特征

# 互信息选择法
from minepy import MINE
import numpy as np
from sklearn import feature_selection

matrix = np.array(X)
target = np.array(target)
def mic(x, y):
    m = MINE()
    m.compute_score(x, y)
    return (m.mic(), 0.5)
temp = feature_selection.SelectKBest(lambda X, Y: np.array(list(map(lambda x: mic(x, Y), X.T))).T[0], k=k).fit(matrix, target)
scores = temp.scores_.tolist()
indx = temp.get_support().tolist()
result = temp.transform(matrix).tolist()
return scores, indx, result

# X: array-like
# target: array-like
# k: int
# http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectKBest.html

5.利用相关系数选择特征

# 利用相关系数选择特征
import numpy as np
from sklearn import feature_selection
from sklearn.feature_selection import chi2

matrix = np.array(X)
target = np.array(target)
temp = feature_selection.SelectKBest(lambda X, Y: np.array(list(map(lambda x: abs(pearsonr(x, Y)[0]), X.T))), k=k).fit(matrix, target)
scores = temp.scores_.tolist()
indx = temp.get_support().tolist()
result = temp.transform(matrix).tolist()
return scores, indx, result

# X: array-like
# target: array-like
# k: int
# http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectKBest.html

6.卡方检验法提取特征

# 卡方检验法提取特征
import numpy as np
from sklearn import feature_selection
from sklearn.feature_selection import chi2

matrix = np.array(X)
target = np.array(target)
temp = feature_selection.SelectKBest(chi2, k=k).fit(matrix, target)
scores = temp.scores_.tolist()
indx = temp.get_support().tolist()
result = temp.transform(matrix).tolist()
return scores, indx, result

# X: array-like
# target: array-like
# k: int
# http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectKBest.html

7.利用方差选择特征

# 利用方差选择特征
import numpy as np
from sklearn import feature_selection

matrix = np.array(X)
temp = feature_selection.VarianceThreshold(threshold=t).fit(matrix)
scores = [np.var(el) for el in matrix.T]
indx = temp.get_support().tolist()
result = temp.transform(matrix).tolist()
return scores, indx, result

# X: array-like
# t: float
# http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.VarianceThreshold.html

参考：https://www.kesci.com/