2

这是一个示例熊猫数据框：

import pandas as pd
import numpy as np

dict1 = {'file': ['filename2', 'filename2', 'filename3', 'filename4', 
         'filename4', 'filename3'], 'amount': [3, 4, 5, 1, 2, 1], 
         'front': [21889611, 36357723, 196312, 11, 42, 1992], 
         'back':[21973805, 36403870, 277500, 19, 120, 3210], 
         'type':['A', 'A', 'A', 'B', 'B', 'C']}

df1 = pd.DataFrame(dict1)
print(df1)

        file  amount     front      back type
0  filename2       3  21889611  21973805    A
1  filename2       4  36357723  36403870    A
2  filename3       5    196312    277500    A
3  filename4       1        11        19    B
4  filename4       2        42       120    B
5  filename3       1      1992      3210    C

我正在end使用以下方法定义一个新列numpy.where()：

df1['end'] = np.where(df1['type']=='B', df1['front'], df1['front'] + df1['back'])
print(df1)
   amount      back       file     front type       end
0       3  21973805  filename2  21889611    A  43863416
1       4  36403870  filename2  36357723    A  72761593
2       5    277500  filename3    196312    A    473812
3       1        19  filename4        11    B        11
4       2       120  filename4        42    B        42
5       1      3210  filename3      1992    C      5202

如果列部分存在，我想使用相同的方法来填充NaN值，例如这里是a确实作为列存在的地方，但有很多endDataFrameendNaN值。（编辑：这些不是 NA 的值可能是完全唯一的）：

new_df
   amount      back       file     front type       end
0       3  21973805  filename2  21889611    A       NaN
1       4  36403870  filename2  36357723    A       NaN
2       5    277500  filename3    196312    A        12
3       1        19  filename4        11    B       NaN
4       2       120  filename4        42    B        49
5       1      3210  filename3      1992    C       NaN

我认为可以使用 来执行此操作pandas.DataFrame.fillna()，但这会引发错误：

df1['end'].fillna(np.where(df1['type']=='B', df1['front'], df1['front'] + df1['back']), inplace=True)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.6/site-packages/pandas/core/series.py", line 2434, in fillna
    **kwargs)
  File "/usr/local/lib/python3.6/site-packages/pandas/core/generic.py", line 3631, in fillna
    type(value))
ValueError: invalid fill value with a <class 'numpy.ndarray'>

问题：我如何有效地np.where()仅使用NaN特定列中的值？

4

1 回答 1

2

fillna基于索引

df['New']=np.where(df1['type']=='B', df1['front'], df1['front'] + df1['back'])
df
Out[125]: 
   amount      back       file     front type       end       New
0       3  21973805  filename2  21889611    A       NaN  43863416
1       4  36403870  filename2  36357723    A       NaN  72761593
2       5    277500  filename3    196312    A  473812.0    473812
3       1        19  filename4        11    B       NaN        11
4       2       120  filename4        42    B      42.0        42
5       1      3210  filename3      1992    C       NaN      5202
df.end.fillna(df.New)
Out[126]: 
0    43863416.0
1    72761593.0
2      473812.0
3          11.0
4          42.0
5        5202.0
Name: end, dtype: float64
df.end=df.end.fillna(df.New)
df
Out[128]: 
   amount      back       file     front type         end       New
0       3  21973805  filename2  21889611    A  43863416.0  43863416
1       4  36403870  filename2  36357723    A  72761593.0  72761593
2       5    277500  filename3    196312    A    473812.0    473812
3       1        19  filename4        11    B        11.0        11
4       2       120  filename4        42    B        42.0        42
5       1      3210  filename3      1992    C      5202.0      5202

更新

df['New']=np.where(df1['type']=='B', df1['front'], df1['front'] + df1['back'])
df.end=df.end.fillna(df.New)
df
Out[133]: 
   amount      back       file     front type         end       New
0       3  21973805  filename2  21889611    A  43863416.0  43863416
1       4  36403870  filename2  36357723    A  72761593.0  72761593
2       5    277500  filename3    196312    A        12.0    473812
3       1        19  filename4        11    B        11.0        11
4       2       120  filename4        42    B        49.0        42
5       1      3210  filename3      1992    C      5202.0      5202
于 2017-12-27T03:18:19.077 回答