示例代码

import pandas as pd
data1 = {'Class': ['Maths', 'English', 'English', 'English', 'Science', 'Science'], 
         'Length': ['Medium', 'Short', 'Long', 'Medium', 'Long', 'Long'], 
         'Head Teacher': ['Mr. Bloggs', 'Mr. Plum', 'Mrs. Green', 'Mr. Top', 'Mrs. Blue', 'Mr. Red'], 
         'Premium Course': ['Yes', 'Yes', 'Yes', 'Yes', 'Yes', 'Yes']}
df = pd.DataFrame(data1)

字符串

第一步

接通条件

m = {'Short':0, 'Medium':1, 'Long':2}
cond = df.groupby('Class')['Length'].transform(lambda x: x.index == x.map(m).idxmax())

型
第二

0     True
1    False
2     True
3    False
4     True
5    False
Name: Length, dtype: bool

型

第二步

编辑列

df['Head Teacher'] = df['Head Teacher'].where(cond).groupby(df['Class']).ffill().bfill()
df['Premium Course'] = df['Premium Course'].where(cond)

型
df

Class   Length  Head Teacher    Premium Course
0   Maths   Medium  Mr. Bloggs      Yes
1   English Short   Mrs. Green      NaN
2   English Long    Mrs. Green      Yes
3   English Medium  Mrs. Green      NaN
4   Science Long    Mrs. Blue       Yes
5   Science Long    Mrs. Blue       NaN

型

使用有序Categorical为Length列，因此可以通过DataFrame.sort_values和DataFrame.duplicated创建掩码，DataFrame.sort_index用于原始行顺序，并将NaN s设置为Series.mask中不匹配的值，其中GroupBy.transform用于获取第一个非NaN值：

df['Length'] = pd.Categorical(df['Length'], 
                              categories=['Long','Medium','Short'],
                              ordered=True)

mask = df.sort_values('Length').duplicated(['Class']).sort_index()
df['Head Teacher'] = df['Head Teacher'].mask(mask).groupby(df['Class']).transform('first')
df['Premium Course'] = df['Premium Course'].mask(mask)

print (df)
     Class  Length Head Teacher Premium Course
0    Maths  Medium   Mr. Bloggs            Yes
1  English   Short   Mrs. Green            NaN
2  English    Long   Mrs. Green            Yes
3  English  Medium   Mrs. Green            NaN
4  Science    Long    Mrs. Blue            Yes
5  Science    Long    Mrs. Blue            NaN

字符串