Numpy 运算对于 groupby 无效。使用 .groupby(...).mean() 代替，这就是我所做的

Question

我想people_preferences根据其段获取以下数据帧的前两列中每一列的平均值Segment。

      Fun|Not-Fun              Pro-garden|Pro-home      Segment
0                          NaN                NaN       cats
1                          NaN                NaN       cats
2                         -1.0                NaN       cats

...                        ...                ...        ...
4570                      -1.0               -1.0       dogs
4571                      -1.0                1.0       dogs
4572                      -1.0                1.0       dogs

所以我尝试了people_preferences.groupby('Segment', as_index=False).mean( skipna = True)，但它返回了：UnsupportedFunctionCall: numpy operations are not valid with groupby. Use .groupby(...).mean() instead

这是完整的错误消息：

---------------------------------------------------------------------------
UnsupportedFunctionCall                   Traceback (most recent call last)
<ipython-input-489-f8da6e73c33c> in <module>
     48 pairs = list(itertools.combinations(df_features.columns, 2))
     49 
---> 50 [plot_mean(pair[0],pair[1]) for pair in pairs]
     51 
     52 fig = px.scatter(df_features, x=columns_x, y=columns_y)

<ipython-input-489-f8da6e73c33c> in <listcomp>(.0)
     48 pairs = list(itertools.combinations(df_features.columns, 2))
     49 
---> 50 [plot_mean(pair[0],pair[1]) for pair in pairs]
     51 
     52 fig = px.scatter(df_features, x=columns_x, y=columns_y)

<ipython-input-489-f8da6e73c33c> in plot_mean(column_x, column_y)
     23         people_preferences = df_features[[column_x,column_y,'Segment']]
     24         print(people_preferences)
---> 25         print(people_preferences.groupby('Segment', as_index=False).mean( skipna = True))
     26         # parties.append('PEOPLE')
     27         dataframe = pd.DataFrame(dict(x=parties_x, y=parties_y, parties = parties))

C:\ProgramData\Anaconda3\lib\site-packages\pandas\core\groupby\groupby.py in mean(self, *args, **kwargs)
   1200         Name: B, dtype: float64
   1201         """
-> 1202         nv.validate_groupby_func("mean", args, kwargs, ["numeric_only"])
   1203         try:
   1204             return self._cython_agg_general(

C:\ProgramData\Anaconda3\lib\site-packages\pandas\compat\numpy\function.py in validate_groupby_func(name, args, kwargs, allowed)
    375                 "numpy operations are not valid "
    376                 "with groupby. Use .groupby(...)."
--> 377                 "{func}() instead".format(func=name)
    378             )
    379         )

Answer 1

这是熊猫中的一个错误。在 groupby 上使用 Skipna查看平均值 True 或 False 给出错误

解决方法

• 使用pandas.aggregate代替

# define helper function
def custom_mean(df):
    return df.mean(skipna=True)

# instead of 
df.mean(skipna=True)

# use 
df.agg(custom_mean)

（注意：可能是 的skipna=True默认值pandas.mean()，尽管由于某种原因文档显示了默认值skipna=None）

Answer 2

我不确定是否正确理解了问题，但是，为了解决问题（不是特别是错误），您应该不会遇到问题：df.groupby(['Segment'])['Fun|Not-Fun','Pro-garden|Pro-home'].mean()因为默认行为是skipna=Truefor mean()。这是一个例子：

import pandas as pd
a = {'a':[1,1,1,2,2,2],'data':[np.nan,10,20,20,30,10],'data_2':[10,20,30,np.nan,10,20]}
df = pd.DataFrame(a)

print(df.groupby('a',as_index=False)['data','data_2'].mean())

输出：

   a  data  data_2
0  1  15.0    20.0
1  2  20.0    15.0