如何基于类别计数排序为Pandas DataFrame创建标签列？

问题描述 #

现有包含category列的Pandas DataFrame，数据如下：

import pandas as pd
data = {'category': ['POLITICS','WELLNESS', 'ENTERTAINMENT', 'TRAVEL','POLITICS', 'ENTERTAINMENT','POLITICS'],
       'dates': ["2013-01-31","2013-01-31","2013-02-02", "2013-02-02","2013-02-03", "2013-02-03", "2013-02-04"]}
df1 = pd.DataFrame(data, columns=['category', 'dates'])

需要生成按类别计数从多到少分配0、1、2...序号的标签字典，用于为DataFrame添加label列。已通过df1["category"].value_counts().to_dict()得到计数字典：{'POLITICS': 3, 'ENTERTAINMENT': 2, 'WELLNESS': 1, 'TRAVEL': 1}

解决方案 #

步骤1：获取按计数降序排列的类别列表 #

value_counts()方法默认按计数降序返回结果，其index属性直接给出排序后的类别序列：

sorted_categories = df1['category'].value_counts().index

步骤2：生成标签字典 #

用enumerate遍历排序后的类别，将每个类别映射为从0开始的整数：

label_dict = {cat: idx for idx, cat in enumerate(sorted_categories)}

生成的字典为：{'POLITICS': 0, 'ENTERTAINMENT': 1, 'WELLNESS': 2, 'TRAVEL': 3}

步骤3：添加label列到DataFrame #

使用map()方法完成标签映射，并用fillna()处理未知类别（示例中默认填充6，可按需调整）：

df1['label'] = df1['category'].map(label_dict).fillna(6).astype(int)

完整代码示例 #

import pandas as pd

# 创建DataFrame
data = {'category': ['POLITICS','WELLNESS', 'ENTERTAINMENT', 'TRAVEL','POLITICS', 'ENTERTAINMENT','POLITICS'],
       'dates': ["2013-01-31","2013-01-31","2013-02-02", "2013-02-02","2013-02-03", "2013-02-03", "2013-02-04"]}
df1 = pd.DataFrame(data, columns=['category', 'dates'])

# 生成按计数排序的标签字典
sorted_categories = df1['category'].value_counts().index
label_dict = {cat: idx for idx, cat in enumerate(sorted_categories)}

# 添加label列
df1['label'] = df1['category'].map(label_dict).fillna(6).astype(int)

# 输出结果
print(df1)

运行后输出的正确结果：

category        dates  label
0        POLITICS  2013-01-31      0
1        WELLNESS  2013-01-31      2
2  ENTERTAINMENT  2013-02-02      1
3          TRAVEL  2013-02-02      3
4        POLITICS  2013-02-03      0
5  ENTERTAINMENT  2013-02-03      1
6        POLITICS  2013-02-04      0

内容的提问来源于stack exchange，提问作者Bluetail