Python特征工程基础

# Python特征工程基础
# 特征工程是将原始数据转换为更适合模型的表示
# 包括缩放、编码、特征构造和特征选择

# 1. 导入库
import numpy as np
import pandas as pd
from sklearn.preprocessing import (
StandardScaler, MinMaxScaler, OneHotEncoder, LabelEncoder, PolynomialFeatures
)
from sklearn.feature_selection import SelectKBest, f_classif
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split

# 2. 创建示例数据（含数值和分类特征）
data = {
'age': [25, 30, 35, 40, 45, 50, 55, 60],
'salary': [30000, 45000, 50000, 60000, 70000, 80000, 90000, 100000],
'city': ['BJ', 'SH', 'GZ', 'BJ', 'SH', 'SZ', 'GZ', 'BJ'],
'department': ['Tech', 'HR', 'Tech', 'Sales', 'Tech', 'HR', 'Sales', 'Tech'],
'target': [0, 0, 1, 1, 1, 0, 0, 1]
}
df = pd.DataFrame(data)
print(f"原始数据:\n{df}\n")

# 3. StandardScaler — 标准化（均值=0，标准差=1）
scaler_std = StandardScaler()
age_salary = df[['age', 'salary']]
age_salary_scaled = scaler_std.fit_transform(age_salary)
print(f"=== StandardScaler ===")
print(f"均值: {age_salary_scaled.mean(axis=0).round(4)}")
print(f"标准差: {age_salary_scaled.std(axis=0).round(4)}")

# 4. MinMaxScaler — 归一化
scaler_mm = MinMaxScaler()
age_salary_mm = scaler_mm.fit_transform(age_salary)
print(f"\n=== MinMaxScaler ===")
print(f"最小值: {age_salary_mm.min(axis=0)}, 最大值: {age_salary_mm.max(axis=0)}")

# 5. OneHotEncoder — 独热编码
ohe = OneHotEncoder(sparse_output=False, drop='first')
city_encoded = ohe.fit_transform(df[['city']])
print(f"\n=== OneHotEncoder ===")
print(f"原始类别: {ohe.categories_}")
print(f"编码结果:\n{city_encoded}")

# 6. LabelEncoder — 标签编码
le = LabelEncoder()
department_encoded = le.fit_transform(df['department'])
print(f"=== LabelEncoder ===")
print(f"编码映射: {dict(zip(le.classes_, le.transform(le.classes_)))}")
print(f"编码结果: {department_encoded}")

# 7. PolynomialFeatures — 多项式特征
poly = PolynomialFeatures(degree=2, include_bias=False)
age_salary_poly = poly.fit_transform(age_salary)
print(f"\n=== PolynomialFeatures ===")
print(f"原始 {age_salary.shape[1]} 维 -> {age_salary_poly.shape[1]} 维")
print(f"特征名称: {poly.get_feature_names_out()}")

# 8. SelectKBest — 特征选择
iris = load_iris()
X, y = iris.data, iris.target
selector = SelectKBest(score_func=f_classif, k=2)
X_selected = selector.fit_transform(X, y)
print(f"\n=== SelectKBest (K=2) ===")
print(f"特征分数: {selector.scores_}")
print(f"选中索引: {selector.get_support(indices=True)}")

# 9. 完整流程：拟合训练集，转换测试集
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.3, random_state=42
)
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test) # 只用 transform！
print(f"\n=== 特征工程总结 ===")
print(f"训练集均值 (≈0): {X_train_scaled.mean(axis=0).round(4)}")

# 10. 特征工程原则
# - 训练集 fit_transform，测试集只 transform（避免数据泄露）
# - 数值型: StandardScaler / MinMaxScaler
# - 无序类别: OneHotEncoder，有序: OrdinalEncoder
# - 特征选择可减少维度、降低过拟合