从基础到高级详解Python中的DataFrame数据转换

import pandas as pd

# 创建示例DataFrame
df = pd.DataFrame({
    'A': [1, 2, 3],
    'B': ['a', 'b', 'c'],
    'C': [1.1, 2.2, 3.3]
})

# 选择单列
col_a = df['A']  # 或 df.A

# 选择多列
subset = df[['A', 'C']]

# 筛选A列大于1的行
filtered = df[df['A'] > 1]

# 多条件筛选
filtered = df[(df['A'] > 1) & (df['C'] < 3)]

# 添加简单计算列
df['D'] = df['A'] * 2

# 添加基于条件的列
df['E'] = ['high' if x > 1 else 'low' for x in df['A']]

# 使用apply函数
df['F'] = df['B'].apply(lambda x: x.upper())

# 填充缺失值
df_filled = df.fillna(0)  # 用0填充
df_filled = df.fillna(df.mean())  # 用均值填充

# 删除包含缺失值的行
df_dropped = df.dropna()

# 字符串转数值
df['A'] = pd.to_numeric(df['A'], errors='coerce')

# 字符串转日期
df['date_col'] = pd.to_datetime(df['date_str'])

# 分类数据转换
df['category_col'] = df['text_col'].astype('category')

from sklearn.preprocessing import MinMaxScaler, StandardScaler

# 归一化 (0-1范围)
scaler = MinMaxScaler()
df[['A', 'C']] = scaler.fit_transform(df[['A', 'C']])

# 标准化 (均值为0，方差为1)
scaler = StandardScaler()
df[['A', 'C']] = scaler.fit_transform(df[['A', 'C']])

# 等宽分箱
df['A_binned'] = pd.cut(df['A'], bins=3, labels=['low', 'medium', 'high'])

# 等频分箱
df['A_qcut'] = pd.qcut(df['A'], q=3, labels=['Q1', 'Q2', 'Q3'])

# 标签编码
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
df['B_encoded'] = le.fit_transform(df['B'])

# 独热编码
df_encoded = pd.get_dummies(df, columns=['B'])

# 创建透 视表
pivot_df = df.pivot_table(index='A', columns='B', values='C', aggfunc='mean')

# 将宽格式转换为长格式
id_vars = ['A']  # 保持不变的列
value_vars = ['C']  # 要熔化的列
melted_df = pd.melt(df, id_vars=id_vars, value_vars=value_vars)

# 堆叠（将多列转换为行）
stacked_df = df.set_index('A').stack().reset_index()

# 解堆叠（将行转换为多列）
unstacked_df = stacked_df.set_index(['A', 'level_1']).unstack().reset_index()

# 创建第二个DataFrame
df2 = pd.DataFrame({
    'A': [1, 2, 4],
    'G': ['x', 'y', 'z']
})

# 内连接
merged_inner = pd.merge(df, df2, on='A', how='inner')

# 左连接
merged_left = pd.merge(df, df2, on='A', how='left')

# 基于索引的连接
left = df.set_index('A')
right = df2.set_index('A')
joined = left.join(right, how='inner')

# 垂直拼接
concatenated = pd.concat([df, df2], axis=0, ignore_index=True)

# 水平拼接
concatenated = pd.concat([df, df2], axis=1)

# 基本分组聚合
grouped = df.groupby('B').agg({
    'A': 'sum',
    'C': ['mean', 'std']
})

# 多级分组
multi_grouped = df.groupby(['B', pd.cut(df['A'], bins=2)]).size()

# 创建时间索引
df['date'] = pd.date_range('2023-01-01', periods=3)
df = df.set_index('date')

# 重采样
monthly_data = df.resample('M').mean()

# 滚动窗口计算
df['rolling_mean'] = df['C'].rolling(window=2).mean()

# 加载数据
sales = pd.read_csv('sales_data.csv')

# 1. 转换日期列
sales['order_date'] = pd.to_datetime(sales['order_date'])

# 2. 提取年月日信息
sales['year'] = sales['order_date'].dt.year
sales['month'] = sales['order_date'].dt.month
sales['day'] = sales['order_date'].dt.day

# 3. 创建价格区间列
bins = [0, 50, 100, 200, float('inf')]
labels = ['0-50', '51-100', '101-200', '200+']
sales['price_range'] = pd.cut(sales['price'], bins=bins, labels=labels)

# 4. 计算每个客户的总消费
customer_spending = sales.groupby('customer_id')['amount'].sum().reset_index()
customer_spending.rename(columns={'amount': 'total_spending'}, inplace=True)

# 5. 合并回原数据集
sales = pd.merge(sales, customer_spending, on='customer_id', how='left')