python - 使用 GroupBy 获取 Pandas 的平均值 - 获取 DataError : No numeric types to aggregate -

Question

我知道有很多关于此的问题，例如 Getting daily averages with pandas和 How get monthly mean in pandas using groupby但我遇到了一个奇怪的错误。

简单的数据集，有一个索引列(时间戳类型)和一个值列。想获得数据的月均值。

In [76]: df.head()
Out[76]: 
                          A
2008-01-02                1
2008-01-03                2
2008-01-04                3
2008-01-07                4
2008-01-08                5

但是，当我进行 groupby 时，我得到的只是索引组而不是值组

In [74]: df.head().groupby(lambda x: x.month).groups
Out[74]: 
{1: [Timestamp('2008-01-02 00:00:00'),
  Timestamp('2008-01-03 00:00:00'),
  Timestamp('2008-01-04 00:00:00'),
  Timestamp('2008-01-07 00:00:00'),
  Timestamp('2008-01-08 00:00:00')]}

尝试采用 means() 会导致错误:

已经尝试了 df.head().resample("M", how='mean') 和 df.head().groupby(lambda x: x.month) .mean()

并得到错误:DataError: No numeric types to aggregate

In [75]: df.resample("M", how='mean')
---------------------------------------------------------------------------
DataError                                 Traceback (most recent call last)
<ipython-input-75-79dc1a060ba4> in <module>()
----> 1 df.resample("M", how='mean')

/usr/local/lib/python2.7/site-packages/pandas/core/generic.pyc in resample(self, rule, how, axis, fill_method, closed, label, convention, kind, loffset, limit, base)
   2878                               fill_method=fill_method, convention=convention,
   2879                               limit=limit, base=base)
-> 2880         return sampler.resample(self).__finalize__(self)
   2881 
   2882     def first(self, offset):

/usr/local/lib/python2.7/site-packages/pandas/tseries/resample.pyc in resample(self, obj)
     82 
     83         if isinstance(ax, DatetimeIndex):
---> 84             rs = self._resample_timestamps()
     85         elif isinstance(ax, PeriodIndex):
     86             offset = to_offset(self.freq)

/usr/local/lib/python2.7/site-packages/pandas/tseries/resample.pyc in _resample_timestamps(self)
    286             # Irregular data, have to use groupby
    287             grouped = obj.groupby(grouper, axis=self.axis)
--> 288             result = grouped.aggregate(self._agg_method)
    289 
    290             if self.fill_method is not None:

/usr/local/lib/python2.7/site-packages/pandas/core/groupby.pyc in aggregate(self, arg, *args, **kwargs)
   2436     def aggregate(self, arg, *args, **kwargs):
   2437         if isinstance(arg, compat.string_types):
-> 2438             return getattr(self, arg)(*args, **kwargs)
   2439 
   2440         result = OrderedDict()

/usr/local/lib/python2.7/site-packages/pandas/core/groupby.pyc in mean(self)
    664         """
    665         try:
--> 666             return self._cython_agg_general('mean')
    667         except GroupByError:
    668             raise

/usr/local/lib/python2.7/site-packages/pandas/core/groupby.pyc in _cython_agg_general(self, how, numeric_only)
   2356 
   2357     def _cython_agg_general(self, how, numeric_only=True):
-> 2358         new_items, new_blocks = self._cython_agg_blocks(how, numeric_only=numeric_only)
   2359         return self._wrap_agged_blocks(new_items, new_blocks)
   2360 

/usr/local/lib/python2.7/site-packages/pandas/core/groupby.pyc in _cython_agg_blocks(self, how, numeric_only)
   2406 
   2407         if len(new_blocks) == 0:
-> 2408             raise DataError('No numeric types to aggregate')
   2409 
   2410         return data.items, new_blocks

DataError: No numeric types to aggregate

Answer 1

是的，您应该尝试使用 df['A'] = df['A'].astype(int) 将 A 强制转换为数字。可能值得检查初始数据读入中是否有任何内容导致它也是对象而不是数字。