如何在Pandas Series中手动合并指定分类？

步骤1：构造示例Series（已有数据可跳过） #

先还原你的数据集为Pandas Series：

import pandas as pd

data = {
    'HS-grad': 15784,
    'Some-college': 10878,
    'Bachelors': 8025,
    'Masters': 2657,
    '11th': 1812,
    '10th': 1389,
    '7th-8th': 955,
    '9th': 756,
    '12th': 657,
    '5th-6th': 509,
    '1st-4th': 247
}
s = pd.Series(data, name='number_of_appearances')

步骤2：自定义合并规则 #

创建映射字典，明确指定哪些分类要合并为新名称，其余分类可映射到自身保持不变：

# 定义合并规则：原分类 → 新分类名
merge_map = {
    '1st-4th': '(1st-4th, 5th-6th)',
    '5th-6th': '(1st-4th, 5th-6th)',
    'HS-grad': 'HS-grad',
    'Some-college': 'Some-college',
    'Bachelors': 'Bachelors',
    'Masters': 'Masters',
    '11th': '11th',
    '10th': '10th',
    '7th-8th': '7th-8th',
    '9th': '9th',
    '12th': '12th'
}

步骤3：执行合并 #

通过map替换分类名，再分组求和得到结果：

merged_s = s.map(merge_map).groupby(level=0).sum()

最终结果 #

运行后得到的merged_s输出如下：

HS-grad               15784
Some-college          10878
Bachelors              8025
Masters                2657
11th                   1812
10th                   1389
7th-8th                 955
9th                     756
12th                    657
(1st-4th, 5th-6th)      756
Name: number_of_appearances, dtype: int64

扩展：多组合并 #

如果需要合并更多分类组，只需修改merge_map即可。比如合并9th、10th、11th、12th为统一名称：

merge_map = {
    '1st-4th': '(1st-4th, 5th-6th)',
    '5th-6th': '(1st-4th, 5th-6th)',
    '9th': 'Lower Secondary/High School Dropout',
    '10th': 'Lower Secondary/High School Dropout',
    '11th': 'Lower Secondary/High School Dropout',
    '12th': 'Lower Secondary/High School Dropout',
    'HS-grad': 'HS-grad',
    'Some-college': 'Some-college',
    'Bachelors': 'Bachelors',
    'Masters': 'Masters',
    '7th-8th': '7th-8th'
}

merged_s = s.map(merge_map).groupby(level=0).sum()

这个方法完全由你自定义合并规则，和R语言forcats包的fct_collapse功能逻辑一致，满足手动指定合并分类的需求。

内容的提问来源于stack exchange，提问作者BoroBoro