Python OneHotEncoder 使用许多虚拟变量或更好的做法？

Question

我正在构建一个神经网络，并且正准备对许多独立(分类)变量使用 OneHotEncoder。我想知道我是否正在使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.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])

这给了我以下数组:

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)

现在我必须为所有自变量创建虚拟变量:

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.]])

或者有更好的方法来解决这个问题吗？

Answer 1

你也可以试试:X=pd.get_dummies(X,columns=[0, 1, 2, 3, 4, 5, 6],drop_first=True)

'drop_first=True' 将您从虚拟变量陷阱中拯救出来。