在使用pandas进行特征工程时,经常会出现空值问题,下面介绍一下pandas中的空值判断
新建一个文件:sample.csv,文件内容如下:
a,b,c,d
1,,1,'1'
2,,2,'2'
3,,,'3'
4,,4,
下面通过pandas的read_csv函数 将文件内容加载到dataframe中,dataFrame的结构如下:
In [3]: data.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 4 entries, 0 to 3
Data columns (total 4 columns):
a 4 non-null int64
b 0 non-null float64
c 3 non-null float64
d 3 non-null object
dtypes: float64(2), int64(1), object(1)
memory usage: 208.0+ bytes
In [4]: data
Out[4]:
a b c d
0 1 NaN 1.0 '1'
1 2 NaN 2.0 '2'
2 3 NaN NaN '3'
3 4 NaN 4.0 NaN
In [5]: data['e'] = None
In [6]: data
Out[6]:
a b c d e
0 1 NaN 1.0 '1' None
1 2 NaN 2.0 '2' None
2 3 NaN NaN '3' None
In [8]: data.isnull()
Out[8]:
a b c d e
0 False True False False True
1 False True False False True
2 False True True False True
3 False True False True True
发现pandas读入的空值默认是NaN,如果赋值为None,则为Nonetype,两者不同,但都可以用DataFrame.isnull()来判断
接下来对空值采取操作,我们的目的是新产生一列数据,这一列数据的值由data['c']或者data['d']来决定,如果某一行的值是空值,则新产生的列的值为1,否则为2。
In [11]: data['f'] = data['c'].apply(lambda x: 1 if x is np.nan else 2)
In [12]: data[['c', 'f']]
Out[12]:
c f
0 1.0 2
1 2.0 2
2 NaN 2
3 4.0 2
In [13]: data['f'] = data['c'].apply(lambda x: 1 if np.isnan(x) else 2)
In [14]: data[['c', 'f']]
Out[14]:
c f
0 1.0 2
1 2.0 2
2 NaN 1
3 4.0 2
In [15]: data['f'] = data['c'].apply(lambda x: 1 if x == np.nan else 2)
In [16]: data[['c', 'f']]
Out[16]:
c f
0 1.0 2
1 2.0 2
2 NaN 2
3 4.0 2
data['c']的数据类型为float,可以发现,使用x is np.nan 或者 x== np.nan 判断的话全部未False,不能正确判断空值,但使用np.isnan(x) 可以达到想要的结果。
In [21]: data['f'] = data['d'].apply(lambda x: 1 if x == np.nan else 2)
In [22]: data[['d', 'f']]
Out[22]:
d f
0 '1' 2
1 '2' 2
2 '3' 2
3 NaN 2
In [23]: data['f'] = data['d'].apply(lambda x: 1 if x is np.nan else 2)
In [24]: data[['d', 'f']]
Out[24]:
d f
0 '1' 2
1 '2' 2
2 '3' 2
3 NaN 1
In [25]: data['f'] = data['d'].apply(lambda x: 1 if np.isnan(x) else 2)
TypeError: ufunc 'isnan' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''
In [30]: data['f'] = data['e'].apply(lambda x: 1 if x is None else 2)
In [31]: data[['e', 'f']]
Out[31]:
e f
0 None 1
1 None 1
2 None 1
3 None 1
In [32]: data['f'] = data['e'].apply(lambda x: 1 if x == None else 2)
In [33]: data[['e', 'f']]
Out[33]:
e f
0 None 1
1 None 1
2 None 1
3 None 1
data['d']的类型为‘object’,发现使用x == np.nan 同样不能识别空值,但是使用x is np.nan则能够识别空值,使用np.isnan(x)则直接报错,因为np.isnan()的输入只能是float
由上可见,对于列的类型为object 的数据类型,判断缺失值使用x is np.nan ,对于float类型: 使用np.isnan(x),对于None判断,可以使用 x is None或者x == None