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。我想知道我是否使用虚拟变量正确处理这个问题,或者由于我的所有变量都需要虚拟变量,因此可能有更好的方法。
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
我导入数据并在自变量上使用 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?