可靠地处理pandas中的nans

Question

什么是最好的，也是最简单的改写NaNs的方法，同时注意不同的情况？

在本例df中，我想根据业务逻辑替换Routed (Expected) Site中的NaNs：

   DBN     DBN - Exam Routed (Expected) Site
00000A  00000A - Scie                 A00000
750000  750000 - Scie                    NaN
840000  840000 - Scie                    NaN

我想要

   DBN     DBN - Exam Routed (Expected) Site
00000A  00000A - Scie                 A00000
750000  750000 - Scie          75 (Hospital)
840000  840000 - Scie           84 (Charter)

我使用了这段代码，它实际上并没有指明NaN是哪种情况：

df['Routed (Expected) Site'].fillna('Charter or Hospital',inplace=True)

我也尝试了一下，不知何故用None覆盖了所有的Routed (Expected) Site

def routed_site_exceptions(DBN, routed_site):
    DBN = str(DBN)
    if DBN.startswith("84") and not routed_site:
        return '84 (Charter)'
    if DBN.startswith('75') and not routed_site:
        return '75 (Hospital)'
df['Routed (Expected) Site'] = np.vectorize(routed_site_exceptions)(df['DBN'],
                                                                    df['Routed (Expected) Site'])

Answer 1

fillna和map

我假设我不知道您的业务逻辑将需要的所有内容。所以我保持了相当的概括性。

def routed_site_exceptions(DBN):
    DBN = str(DBN)
    if DBN.startswith("84"):
        return '84 (Charter)'
    if DBN.startswith('75'):
        return '75 (Hospital)'

df.fillna({'Routed (Expected) Site': df.DBN.map(routed_site_exceptions)})

      DBN     DBN - Exam Routed (Expected) Site
0  00000A  00000A - Scie                 A00000
1  750000  750000 - Scie          75 (Hospital)
2  840000  840000 - Scie           84 (Charter)

Answer 2

如上所述，您可以执行以下操作：

# get the new values
s = df.DBN.str[:2].map({'75': '75 (Hospital)',
                        '84': '84 (Charter)'})

# map to the data
df['Routed (Expected) Site'].fillna(s, inplace=True)

输出：

      DBN     DBN - Exam Routed (Expected) Site
0  00000A  00000A - Scie                 A00000
1  750000  750000 - Scie          75 (Hospital)
2  840000  840000 - Scie           84 (Charter)