在管道中对目标进行标签编码

Question

我想创建一个管道来对训练特征和目标进行预处理，然后训练模型。数据集类似于：

   v1  v2  target
0  1   a   yes
1  5   c   no
2  3   f   yes

我有一个类似的管道

num_cols = ['v1']
cat_cols = ['v2']
clf = DecisionTreeClassifier()

num_transformer = Pipeline(steps=[
    ('impute', SimpleImputer(strategy='mean')),
    ('scale', MinMaxScaler())
])
cat_transformer = Pipeline(steps=[('impute', SimpleImputer(strategy='most_frequent'))])

col_trans = ColumnTransformer(transformers=[
    ('num_pipeline', num_transformer, num_cols),
    ('cat_pipeline', cat_transformer, cat_cols))

clf_pipeline = Pipeline(steps=[
    ('col_trans', col_trans),
    ('model', clf)
])

这个想法是对目标进行标签编码，这样它就会是

target
1
0
1

如果可能的话，解码预测也会很有趣

Answer 1

解决此问题的一种方法如下：

import pandas as pd
from sklearn.pipeline import Pipeline
from sklearn.compose import ColumnTransformer
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import MinMaxScaler, OneHotEncoder, LabelEncoder
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

data = {
    'v1': [1, 5, 3],
    'v2': ['a', 'c', 'f'],
    'target': ['yes', 'no', 'yes']
}
df = pd.DataFrame(data)

X = df.drop('target', axis=1)
y = df['target']

num_cols = ['v1']
cat_cols = ['v2']

num_transformer = Pipeline(steps=[
    ('impute', SimpleImputer(strategy='mean')),
    ('scale', MinMaxScaler())
])

cat_transformer = Pipeline(steps=[
    ('impute', SimpleImputer(strategy='most_frequent')),
    ('onehot', OneHotEncoder(handle_unknown='ignore'))
])

col_trans = ColumnTransformer(transformers=[
    ('num_pipeline', num_transformer, num_cols),
    ('cat_pipeline', cat_transformer, cat_cols)
])

clf = DecisionTreeClassifier()
clf_pipeline = Pipeline(steps=[
    ('col_trans', col_trans),
    ('model', clf)
])

le = LabelEncoder()
y_encoded = le.fit_transform(y)

X_train, X_test, y_train, y_test = train_test_split(X, y_encoded, test_size=0.2, random_state=42)

clf_pipeline.fit(X_train, y_train)
y_pred = clf_pipeline.predict(X_test)
y_pred_decoded = le.inverse_transform(y_pred)

accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy}')
print(f'Predictions: {y_pred_decoded}')

Wich 给出编码

array([1, 0, 1])

和预测

Accuracy: 1.0
Predictions: ['yes']