[英]DataFrame filter based on groupby
Here is my simplified example df: 这是我的简化示例df:
salesPerson customer measure timeStamp
--------------------------------------
A 123 I 12:30
A 123 II 12:30
A 123 III 12:30
B 123 IV 12:35
C 456 I 14:30
C 456 II 14:30
D 456 III 14:15
What I want to do, it to filer the dataframe and in cases when 2 different salesPerson Id's have the same customer number, keep all the rows of the salesPerson whos timeStamp is the earliest. 我要执行的操作是归档数据帧,并在2个不同的salesPerson ID具有相同客户编号的情况下,请保留timeStamp最早的salesPerson的所有行。 Resulting df in this example would be: 在此示例中,结果df为:
salesPerson customer measure timeStamp
--------------------------------------
A 123 I 12:30
A 123 II 12:30
A 123 III 12:30
D 456 III 14:15
What would be the best/most pythonic way to do it? 最好/最有效的方法是什么? I thought about using pandas groupby.filter or groupby.transform, but frankly have no idea how to accurately write those. 我考虑过使用熊猫groupby.filter或groupby.transform,但是坦率地说,不知道如何准确地编写它们。
Bonus points would be for having the deleted rows in a separate deleted_df object. 奖励点是将删除的行放在单独的Deleted_df对象中。
This one-liner should do the trick: 这种单线应该可以解决问题:
df[df['salesPerson'].isin(df.iloc[df.groupby(['customer'])['timeStamp'].idxmin(), 'salesPerson'])]
Explanation: 说明:
To determine the salespersons to whom we want to filter, first group df
by customer
and get the index where the minimum timeStamp
is found using idxmin
: 为了确定我们要过滤的销售人员,首先按customer
对df
进行分组,并使用idxmin
timeStamp
找到最小timeStamp
的索引:
df.groupby(['customer'])['timeStamp'].idxmin()
Then, pass those index values to iloc
, along with the column we want, to get the values from salesPerson
we'll use for filtering: 然后,将这些索引值以及iloc
,以从用于过滤的salesPerson
获取值:
df.iloc[df.groupby(['customer'])['timeStamp'].idxmin(), 'salesPerson']
Finally, pass that result to the Series method isin
, and use that to index into df
. 最后,将该结果传递给Series方法isin
,并使用该结果索引到df
。 The result is thus: 结果是:
0 A 123 I 2017-07-12 12:30:00
1 A 123 II 2017-07-12 12:30:00
2 A 123 III 2017-07-12 12:30:00
6 D 456 III 2017-07-12 14:15:00
To create a second DataFrame with the filtered-out rows, you could pass the index from the filtered df to the original df and exclude those rows. 要创建带有已过滤出行的第二个DataFrame,可以将索引从已过滤df传递到原始df,并排除这些行。 So if we assigned the result above to df1
, we could create a complementary df2
in this manner: 因此,如果我们将上述结果分配给df1
,则可以按以下方式创建互补的df2
:
df2 = df[~df.index.isin(df1.index)]
Result: 结果:
3 B 123 IV 2017-07-12 12:35:00
4 C 456 I 2017-07-12 14:30:00
5 C 456 II 2017-07-12 14:30:00
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.