熊猫：DataFrame.mean（）非常慢。如何更快地计算列均值？

这与的大小相似，但没有对象列

In [10]: nrows = 10000000In [11]: df = pd.concat([Dataframe(randn(int(nrows),34),columns=[ 'f%s' % i for i in range(34) ]),Dataframe(randint(0,10,size=int(nrows*19)).reshape(int(nrows),19),columns=[ 'i%s' % i for i in range(19) ])],axis=1)In [12]: df.iloc[1000:10000,0:20] = np.nanIn [13]: df.info()<class 'pandas.core.frame.Dataframe'>Int64Index: 10000000 entries, 0 to 9999999Data columns (total 53 columns):f0     9991000  non-null valuesf1     9991000  non-null valuesf2     9991000  non-null valuesf3     9991000  non-null valuesf4     9991000  non-null valuesf5     9991000  non-null valuesf6     9991000  non-null valuesf7     9991000  non-null valuesf8     9991000  non-null valuesf9     9991000  non-null valuesf10    9991000  non-null valuesf11    9991000  non-null valuesf12    9991000  non-null valuesf13    9991000  non-null valuesf14    9991000  non-null valuesf15    9991000  non-null valuesf16    9991000  non-null valuesf17    9991000  non-null valuesf18    9991000  non-null valuesf19    9991000  non-null valuesf20    10000000  non-null valuesf21    10000000  non-null valuesf22    10000000  non-null valuesf23    10000000  non-null valuesf24    10000000  non-null valuesf25    10000000  non-null valuesf26    10000000  non-null valuesf27    10000000  non-null valuesf28    10000000  non-null valuesf29    10000000  non-null valuesf30    10000000  non-null valuesf31    10000000  non-null valuesf32    10000000  non-null valuesf33    10000000  non-null valuesi0     10000000  non-null valuesi1     10000000  non-null valuesi2     10000000  non-null valuesi3     10000000  non-null valuesi4     10000000  non-null valuesi5     10000000  non-null valuesi6     10000000  non-null valuesi7     10000000  non-null valuesi8     10000000  non-null valuesi9     10000000  non-null valuesi10    10000000  non-null valuesi11    10000000  non-null valuesi12    10000000  non-null valuesi13    10000000  non-null valuesi14    10000000  non-null valuesi15    10000000  non-null valuesi16    10000000  non-null valuesi17    10000000  non-null valuesi18    10000000  non-null valuesdtypes: float64(34), int64(19)

时间（与您相似的机器规格）

In [14]: %timeit df.mean()1 loops, best of 3: 21.5 s per loop

您可以通过预先转换为浮点数来获得2倍的加速（平均值是这样做的，但是这样做的方式更通用，所以更慢）

In [15]: %timeit df.astype('float64').mean()1 loops, best of 3: 9.45 s per loop

您的问题是对象列。Mean将尝试为所有列进行计算，但是由于对象列的存在，所有内容都被转换为

object

最好的选择是做

 df._get_numeric_data().mean()

可以

numeric_only

False