PythonOneHotEncoder使用许多虚拟变量还是更好的实践?
我正在构建一个神经网络,并且正在对许多独立(分类)变量使用OneHotEncoder。我想知道我是否用虚拟变量正确地处理这个问题,或者因为我的所有变量都需要虚拟变量,所以可能有更好的方法PythonOneHotEncoder使用许多虚拟变量还是更好的实践?,python,neural-network,dummy-variable,one-hot-encoding,Python,Neural Network,Dummy Variable,One Hot Encoding,我正在构建一个神经网络,并且正在对许多独立(分类)变量使用OneHotEncoder。我想知道我是否用虚拟变量正确地处理这个问题,或者因为我的所有变量都需要虚拟变量,所以可能有更好的方法 df UserName Token ThreadID ChildEXE 0 TAG TokenElevationTypeDefault (1) 20788 splunk-MonitorNoHandle.
df
UserName Token ThreadID ChildEXE
0 TAG TokenElevationTypeDefault (1) 20788 splunk-MonitorNoHandle.exe
1 TAG TokenElevationTypeDefault (1) 19088 splunk-optimize.exe
2 TAG TokenElevationTypeDefault (1) 2840 net.exe
807 User TokenElevationTypeFull (2) 18740 E2CheckFileSync.exe
808 User TokenElevationTypeFull (2) 18740 E2check.exe
809 User TokenElevationTypeFull (2) 18740 E2check.exe
811 Local TokenElevationTypeFull (2) 18740 sc.exe
ParentEXE ChildFilePath ParentFilePath
splunkd.exe C:\Program Files\Splunk\bin C:\Program Files\Splunk\bin 0
splunkd.exe C:\Program Files\Splunk\bin C:\Program Files\Splunk\bin 0
dagent.exe C:\Windows\System32 C:\Program Files\Dagent 0
wscript.exe \Device\Mup\sysvol C:\Windows 1
E2CheckFileSync.exe C:\Util \Device\Mup\sysvol\ 1
cmd.exe C:\Windows\SysWOW64 C:\Util\E2Check 1
cmd.exe C:\Windows C:\Windows\SysWOW64 1
DependentVariable
0
0
0
1
1
1
1
我导入数据并对自变量使用LabelEncoder
from sklearn.preprocessing import LabelEncoder, OneHotEncoder
#IMPORT DATA
#Matrix x of features
X = df.iloc[:, 0:7].values
#Dependent variable
y = df.iloc[:, 7].values
#Encoding Independent Variable
#Need a label encoder for every categorical variable
#Converts categorical into number - set correct index of column
#Encode "UserName"
labelencoder_X_1 = LabelEncoder()
X[:, 0] = labelencoder_X_1.fit_transform(X[:, 0])
#Encode "Token"
labelencoder_X_2 = LabelEncoder()
X[:, 1] = labelencoder_X_2.fit_transform(X[:, 1])
#Encode "ChildEXE"
labelencoder_X_3 = LabelEncoder()
X[:, 3] = labelencoder_X_3.fit_transform(X[:, 3])
#Encode "ParentEXE"
labelencoder_X_4 = LabelEncoder()
X[:, 4] = labelencoder_X_4.fit_transform(X[:, 4])
#Encode "ChildFilePath"
labelencoder_X_5 = LabelEncoder()
X[:, 5] = labelencoder_X_5.fit_transform(X[:, 5])
#Encode "ParentFilePath"
labelencoder_X_6 = LabelEncoder()
X[:, 6] = labelencoder_X_6.fit_transform(X[:, 6])
这为我提供了以下数组:
X
array([[2, 0, 20788, ..., 46, 31, 24],
[2, 0, 19088, ..., 46, 31, 24],
[2, 0, 2840, ..., 27, 42, 15],
...,
[2, 0, 20148, ..., 17, 40, 32],
[2, 0, 20148, ..., 47, 23, 0],
[2, 0, 3176, ..., 48, 42, 32]], dtype=object)
现在,对于所有自变量,我必须创建虚拟变量:
我应该使用:
onehotencoder = OneHotEncoder(categorical_features = [0, 1, 2, 3, 4, 5, 6])
X = onehotencoder.fit_transform(X).toarray()
这给了我:
X
array([[0., 0., 1., ..., 0., 0., 0.],
[0., 0., 1., ..., 0., 0., 0.],
[0., 0., 1., ..., 0., 0., 0.],
...,
[0., 0., 1., ..., 1., 0., 0.],
[0., 0., 1., ..., 0., 0., 0.],
[0., 0., 1., ..., 1., 0., 0.]])
还是有更好的方法来解决这个问题?这是我能找到的最好的方法,而且效果很好:
onehotencoder = OneHotEncoder(categorical_features = [0,1,2,3,4,5,6])
X = onehotencoder.fit_transform(X).toarray()
您也可以尝试:
X=pd.get_假人(X,列=[0,1,2,3,4,5,6],drop_first=True)
“drop_first=True”将您从虚拟变量陷阱中解救出来