Question

我有一个包含毫安行和大量NaN值的DataFrame。一些例子：

index     Company        Area
    0     Google         Technology
    1     Coca Cola      Drinks
    2     NaN            Drinks
    3     Apple          Technology
    4     NaN            Technology
    5     Gatorade       Drinks
    6     Dell           Technology
    7     Apple          Technology
    8     Coca Cola      Drinks
    9     NaN            Drinks
    10    Google         Technology

我的想法是使用其区域的2个最常见值之一填充公司NaN值。

从示例：如果技术领域最常见的公司是Apple和Google，我想填写“df ['Area'] =='技术'”NaN值与其中一个值（随机）

我已经创建了一个具有最常见值的Group By DataFrame，它是这样的：

Area          Company
Technology    Google
Technology    Apple
Drinks        Coca Cola
Drinks        Pepsi

结果应该是这样的：

index     Company        Area
    0     Google         Technology
    1     Coca Cola      Drinks
    2     Pepsi          Drinks
    3     Apple          Technology
    4     Google         Technology
    5     Gatorade       Drinks
    6     Dell           Technology
    7     Apple          Technology
    8     Coca Cola      Drinks
    9     Pepsi          Drinks
    10    Google         Technology

我希望你能帮助我。

感谢!!!

Answer 1

我使用random.choice

来提出这个解决方案

import random

s=df1.groupby('Area').Company.apply(list).reindex(df.Area).apply(lambda x :random.choice(x) )
s.index=df.index

df.Company=df.Company.fillna(s)

df
Out[200]: 
    index   Company        Area
0       0    Google  Technology
1       1  CocaCola      Drinks
2       2  CocaCola      Drinks
3       3     Apple  Technology
4       4    Google  Technology
5       5  Gatorade      Drinks
6       6      Dell  Technology
7       7     Apple  Technology
8       8  CocaCola      Drinks
9       9     Pepsi      Drinks
10     10    Google  Technology

Answer 2

设置

import io

z=io.StringIO("""
Company        Area
Google         Technology
CocaCola      Drinks
NaN            Drinks
Apple          Technology
NaN            Technology
Gatorade       Drinks
Dell           Technology
Apple          Technology
CocaCola      Drinks
NaN            Drinks
Google         Technology""")

df = pd.read_table(z, delim_whitespace=True)

然后你可以做

t = df.groupby("Area").Company.value_counts()
s = t.groupby("Area").apply(lambda x: [(i[1]) for i,v in zip(x.index,x) if v==max(x)])

其中s是具有最常见值的系列。例如：

>>> s
Area
Drinks             [CocaCola]
Technology    [Apple, Google]
Name: Company, dtype: object

现在使用random.choice

from random import choice
df2 = df.set_index("Area")
mask = df2.Company.isna()
df2.loc[mask, "Company"] = [choice(s[s.index == i].item()) for i in df2.loc[mask].index]

使用来自另一个DataFrame pandas的随机值填充nan值

2 个答案:

设置