missing data
pandas
对象上的所有描述统计都排除了缺失数据。
pandas
用NaN(Not a Number)表示浮点数和非浮点数组中的缺失数据,它只是一个便于检测的标记而已。
滤除缺失数据(dropna):
dropna
:
对于一个Series
,dropna
返回一个仅含非空数据和索引值的Series
:
In [55]: from numpy import nan as NA
In [56]: data = Series([1, NA, 3.5, NA, 7])
In [57]: data.dropna()
Out[57]:
0 1.0
2 3.5
4 7.0
dtype: float64
而对于DataFrame
对象,情况有点复杂:dropna
默认丢弃任何含有缺失的行:
In [58]: data = DataFrame([[1., 6.5, 3.], [1., NA, NA], [NA, NA, NA], [NA, 6.5, 3.]])
In [59]: cleaned = data.dropna()
In [60]: data
Out[60]:
0 1 2
0 1.0 6.5 3.0
1 1.0 NaN NaN
2 NaN NaN NaN
3 NaN 6.5 3.0
In [62]: cleaned
Out[62]:
0 1 2
0 1.0 6.5 3.0
传入 how = 'all'
将只丢弃全部为NA
的行:
In [63]: data.dropna(how='all')
Out[63]:
0 1 2
0 1.0 6.5 3.0
1 1.0 NaN NaN
3 NaN 6.5 3.0
丢弃列传入axis=1
即可
填补缺失数据(fillna):
对于大多数情况而言,fillna
方法是主要的函数:
通过一个常数调用fillna
就会将缺失值替换为那个常数值:
In [70]: df.fillna(0)
Out[70]:
0 1 2
0 1.399904 0.000000 0.000000
1 1.575537 0.000000 0.000000
2 -0.331882 0.000000 -0.649257
3 0.098962 0.000000 -0.488681
4 0.816113 0.578439 -0.255484
5 -0.794037 -0.190817 -1.670218
6 -0.319344 0.112990 1.457793
如果通过一个字典调用fillna
,就可以实现对不同的列填充不同的值:
In [71]: df.fillna({1:0.5, 3:-1})
Out[71]:
0 1 2
0 1.399904 0.500000 NaN
1 1.575537 0.500000 NaN
2 -0.331882 0.500000 -0.649257
3 0.098962 0.500000 -0.488681
4 0.816113 0.578439 -0.255484
5 -0.794037 -0.190817 -1.670218
6 -0.319344 0.112990 1.457793
fillna
默认返回新对象,也可以对现有对象进行就地修改:
In [72]: _ = df.fillna(0, inplace=True)
In [73]: df
Out[73]:
0 1 2
0 1.399904 0.000000 0.000000
1 1.575537 0.000000 0.000000
2 -0.331882 0.000000 -0.649257
3 0.098962 0.000000 -0.488681
4 0.816113 0.578439 -0.255484
5 -0.794037 -0.190817 -1.670218
6 -0.319344 0.112990 1.457793
对reindex
有效的那些插值方法也可用于fillna
:
In [75]: df = DataFrame(np.random.randn(6,3))
In [76]: df.iloc[:2, 1] = NA; df.iloc[4:, 2] = NA
In [77]: df
Out[77]:
0 1 2
0 0.877356 NaN -1.775499
1 -0.599936 NaN 0.891599
2 -0.234968 -0.438411 1.519332
3 -1.026612 0.409573 0.667059
4 -1.491810 -0.316408 NaN
5 -0.185388 0.778041 NaN
In [78]: df.fillna(method='ffill')
Out[78]:
0 1 2
0 0.877356 NaN -1.775499
1 -0.599936 NaN 0.891599
2 -0.234968 -0.438411 1.519332
3 -1.026612 0.409573 0.667059
4 -1.491810 -0.316408 0.667059
5 -0.185388 0.778041 0.667059
In [79]: df.fillna(method='ffill', limit=2)
Out[79]:
0 1 2
0 0.877356 NaN -1.775499
1 -0.599936 NaN 0.891599
2 -0.234968 -0.438411 1.519332
3 -1.026612 0.409573 0.667059
4 -1.491810 -0.316408 0.667059
5 -0.185388 0.778041 0.667059
网友评论