python - 将 pandas 数据帧映射到具有组合键的字典

Question

我正在编写一些代码，需要将 pandas 数据帧映射到由复合键和某个值组成的字典中。下面是一个起始示例，键由(PostalCode, Sex)(Name, Age)组成，value是 > 与key 匹配的所有salary 的总和。我正在寻找一种优雅地完成此映射的方法。

import pandas as pd

data = [
    ["tom", 22, "ab 11", "M", 5555],
    ["Rob", 22, "ab 13", "M", 9999],
    ["nick", 33, "ab 14", "M", 3333],
    ["nick", 33, "ab 14", "M", 8888],
    ["juli", 18, "ab 15", "F", 2222],
]
people = pd.DataFrame(data, columns=["Name", "Age", "PostalCode", "Sex", "Salary"])

df = people.groupby(["PostalCode", "Sex", "Age"])["Salary"].sum().unstack(0)

d = {col: df[col].dropna().to_dict() for col in df}

print(d)

# Expected output
print(
    {
        (("ab 11", "M"), ("tom", 22)): 5555,
        (("ab 13", "M"), ("Rob", 22)): 9999,
        (("ab 14", "M"), ("nick", 33)): 12221,
        (("ab 15", "F"), ("juli", 18)): 2222,
    }
)

Answer 1

首先聚合sum，然后在字典理解中更改MultiIndex中值的格式，并将键解包为变量a,b,c,d :

s = people.groupby(["PostalCode", "Sex","Name", "Age"])["Salary"].sum()
print (s)
PostalCode  Sex  Name  Age
ab 11       M    tom   22      5555
ab 13       M    Rob   22      9999
ab 14       M    nick  33     12221
ab 15       F    juli  18      2222
Name: Salary, dtype: int64

d= {((a,b), (c,d)): v for (a,b,c,d), v in s.items()}
print(d)
{(('ab 11', 'M'), ('tom', 22)): 5555, 
 (('ab 13', 'M'), ('Rob', 22)): 9999, 
 (('ab 14', 'M'), ('nick', 33)): 12221, 
 (('ab 15', 'F'), ('juli', 18)): 2222}