如何对数据集列做One Hot Encoding？pd.get_dummies报错求助

原始数据集：

bread   milk    butter  jam  nutella    cheese  chips
0   bread   NaN     butter  jam  nutella    NaN     NaN
1   NaN     NaN     butter  jam  nutella    NaN     chips
2   NaN     milk    NaN     NaN  NaN        cheese  NaN
3   bread   milk    butter  jam  nutella    cheese  chips
4   bread   milk    NaN     NaN  nutella    NaN     NaN
5   bread   milk    butter  jam  NaN        cheese  chips
6   bread   milk    NaN     NaN  nutella    NaN     NaN
7   bread   NaN     butter  NaN  NaN        cheese  NaN
8   bread   NaN     butter  jam  nutella    NaN     NaN
9   NaN     milk    butter  jam  NaN        cheese  NaN
10  bread   NaN     NaN     jam  nutella    cheese  chips
11  bread   milk    butter  jam  nutella    NaN     NaN
12  bread   NaN     butter  NaN  nutella    cheese  NaN
13  bread   NaN     butter  jam  nutella    cheese  chips
14  bread   milk    butter  jam  nutella    cheese  chips
15  NaN     milk    butter  jam  nutella    cheese  NaN
16  NaN     milk    NaN     jam  nutella    cheese  NaN
17  bread   milk    butter  jam  nutella    cheese  chips
18  bread   NaN     butter  jam  nutella    cheese  NaN
19  bread   milk    butter  NaN  nutella    cheese  NaN
20  NaN     milk    NaN     NaN  NaN        NaN     chips

需求：对每列进行One Hot Encoding，生成如下格式的全量0/1数据集：

尝试代码：

pd.get_dummies(book_data, columns = ['bread', 'milk','butter','jam', 'nutella','cheese','chips'])

出现错误：

KeyError: "['bread', 'cheese'] not in index"

你用错了方法——pd.get_dummies是给分类变量生成新的独热编码列的工具，但你的需求是直接把每一列转换成「是否存在该商品」的0/1值，而非对分类值编码。另外报错的核心原因是原始数据的列名可能带有前后空格，导致代码找不到对应列。

正确实现代码：

import pandas as pd

# 假设原始数据已读取为DataFrame对象df
# 第一步：去除列名前后的空格，解决KeyError问题
df.columns = df.columns.str.strip()

# 第二步：将每列转换为0/1：非空值（存在商品）为1，空值（NaN）为0
one_hot_df = df.notna().astype(int)

# 查看结果
print(one_hot_df)

这段代码完成两个关键操作：

1. 清理列名，消除因空格导致的列名匹配失败
2. 通过notna()判断单元格是否存在商品，将布尔结果转为整数，直接得到目标格式的0/1数据集

内容的提问来源于stack exchange，提问作者Addy