本文目录导读：
我来详细介绍Pandas进行数据归一化的几种常用方法。
准备示例数据
import pandas as pd
import numpy as np
# 创建示例数据
data = {
'年龄': [25, 30, 35, 40, 45, 50],
'收入': [50000, 60000, 75000, 80000, 120000, 150000],
'消费分数': [60, 70, 80, 90, 95, 85]
}
df = pd.DataFrame(data)
print("原始数据：")
print(df)
Min-Max归一化（0-1区间）
def min_max_normalize(df, columns=None):
"""
Min-Max归一化：将数据缩放到[0,1]区间
公式: X_norm = (X - X_min) / (X_max - X_min)
"""
if columns is None:
columns = df.columns
df_normalized = df.copy()
for col in columns:
min_val = df[col].min()
max_val = df[col].max()
df_normalized[col] = (df[col] - min_val) / (max_val - min_val)
return df_normalized
# 应用Min-Max归一化
df_minmax = min_max_normalize(df, ['年龄', '收入', '消费分数'])
print("\nMin-Max归一化结果：")
print(df_minmax.round(3))
Z-Score标准化（均值为0，标准差为1）
def z_score_normalize(df, columns=None):
"""
Z-Score标准化：使数据服从标准正态分布
公式: X_std = (X - μ) / σ
"""
if columns is None:
columns = df.columns
df_normalized = df.copy()
for col in columns:
mean_val = df[col].mean()
std_val = df[col].std()
df_normalized[col] = (df[col] - mean_val) / std_val
return df_normalized
# 应用Z-Score标准化
df_zscore = z_score_normalize(df, ['年龄', '收入', '消费分数'])
print("\nZ-Score标准化结果：")
print(df_zscore.round(3))
使用sklearn进行归一化
from sklearn.preprocessing import MinMaxScaler, StandardScaler, RobustScaler
# 方法1：Min-Max归一化
scaler_minmax = MinMaxScaler()
df_sklearn_minmax = pd.DataFrame(
scaler_minmax.fit_transform(df[['年龄', '收入', '消费分数']]),
columns=['年龄_minmax', '收入_minmax', '消费分数_minmax']
)
print("\nSklearn Min-Max归一化：")
print(df_sklearn_minmax.round(3))
# 方法2：Z-Score标准化
scaler_standard = StandardScaler()
df_sklearn_standard = pd.DataFrame(
scaler_standard.fit_transform(df[['年龄', '收入', '消费分数']]),
columns=['年龄_std', '收入_std', '消费分数_std']
)
print("\nSklearn Z-Score标准化：")
print(df_sklearn_standard.round(3))
# 方法3：Robust缩放（对异常值鲁棒）
scaler_robust = RobustScaler()
df_sklearn_robust = pd.DataFrame(
scaler_robust.fit_transform(df[['年龄', '收入', '消费分数']]),
columns=['年龄_robust', '收入_robust', '消费分数_robust']
)
print("\nRobust缩放（基于中位数和IQR）：")
print(df_sklearn_robust.round(3))
自定义归一化函数
def custom_normalize(df, method='minmax', columns=None):
"""
自定义归一化函数，支持多种方法
"""
if columns is None:
columns = df.columns
df_normalized = df.copy()
for col in columns:
if method == 'minmax':
# Min-Max归一化
min_val = df[col].min()
max_val = df[col].max()
df_normalized[col] = (df[col] - min_val) / (max_val - min_val)
elif method == 'zscore':
# Z-Score标准化
mean_val = df[col].mean()
std_val = df[col].std()
df_normalized[col] = (df[col] - mean_val) / std_val
elif method == 'maxabs':
# 最大绝对值归一化
max_abs = df[col].abs().max()
df_normalized[col] = df[col] / max_abs
elif method == 'median':
# 中位数归一化
median_val = df[col].median()
iqr_val = df[col].quantile(0.75) - df[col].quantile(0.25)
df_normalized[col] = (df[col] - median_val) / iqr_val
return df_normalized
# 使用自定义函数
df_custom = custom_normalize(df, method='minmax', columns=['年龄', '收入'])
print("\n自定义归一化结果：")
print(df_custom.round(3))
实用案例：混合数据处理
def normalize_mixed_data(df, numeric_columns, method='zscore'):
"""
处理混合类型数据的归一化
"""
df_normalized = df.copy()
# 仅对数值列进行归一化
for col in numeric_columns:
if method == 'zscore':
mean_val = df[col].mean()
std_val = df[col].std()
df_normalized[f'{col}_normalized'] = (df[col] - mean_val) / std_val
elif method == 'minmax':
min_val = df[col].min()
max_val = df[col].max()
df_normalized[f'{col}_normalized'] = (df[col] - min_val) / (max_val - min_val)
return df_normalized
# 创建包含混合类型的数据
mixed_data = {
'姓名': ['张三', '李四', '王五', '赵六'],
'年龄': [25, 30, 35, 40],
'收入': [50000, 60000, 75000, 80000],
'性别': ['男', '女', '男', '女'],
'消费分数': [60, 70, 80, 90]
}
df_mixed = pd.DataFrame(mixed_data)
# 对数值列进行归一化
df_mixed_normalized = normalize_mixed_data(df_mixed, ['年龄', '收入', '消费分数'], method='zscore')
print("\n混合数据归一化结果：")
print(df_mixed_normalized.round(3))
批量处理并保存结果
def batch_normalize_and_save(df, normalize_methods=['minmax', 'zscore', 'robust']):
"""
批量应用不同的归一化方法并保存结果
"""
results = {}
for method in normalize_methods:
if method == 'minmax':
normalized_df = custom_normalize(df, method='minmax')
elif method == 'zscore':
normalized_df = custom_normalize(df, method='zscore')
elif method == 'robust':
normalized_df = custom_normalize(df, method='median')
results[method] = normalized_df
return results
# 批量处理
all_normalizations = batch_normalize_and_save(df[['年龄', '收入', '消费分数']])
# 显示每种方法的结果
for method, normalized_df in all_normalizations.items():
print(f"\n{method.upper()} 归一化：")
print(normalized_df.round(3))
选择建议
- Min-Max归一化：适用于有明确边界的场景，如像素值[0,255]
- Z-Score标准化：适用于数据近似正态分布，需处理异常值
- Robust缩放：数据包含异常值时更稳定
- MaxAbs归一化：稀疏数据或已有正负值的数据
这些方法可以根据具体需求灵活选择和应用！
