[英]Pandas Group By then Merge Using Multiple Columns
I'm pretty new to Pandas but I'm trying to analyze a dataset of employee timestamps to determine the sum of unique daily timestamps per week. 我是Pandas的新手,但我试图分析员工时间戳的数据集,以确定每周唯一的每日时间戳的总和。
My initial dataframe (input1) looks like this (but much longer): 我的初始数据帧(input1)看起来像这样(但更长):
ID Datetime Week/Year
0 15.0 2019-02-04 08:28:44 6/2019
1 15.0 2019-02-04 12:48:05 6/2019
2 15.0 2019-02-04 12:54:29 6/2019
3 15.0 2019-02-05 08:05:51 6/2019
4 15.0 2019-02-05 12:47:26 6/2019
5 15.0 2019-02-05 14:45:34 6/2019
6 15.0 2019-02-06 08:10:59 6/2019
7 15.0 2019-02-06 12:49:24 6/2019
8 15.0 2019-02-06 13:02:48 6/2019
9 15.0 2019-02-07 08:02:22 6/2019
10 15.0 2019-02-08 08:02:10 6/2019
11 15.0 2019-02-08 09:55:22 6/2019
I created another dataframe: 我创建了另一个数据框:
df = pd.DataFrame({'Timestamp': input1['Datetime'], 'ID': input1['ID'], 'Week/Year': input1['Week/Year'],'MDY':input1['Server Date/Time'].apply(lambda x: "%d/%d/%d" % (x.month, x.day, x.year))})
Then I grouped by Week, Employee, and got unique count per day (MDY): 然后,我按周,员工分组,并获得每天的唯一计数(MDY):
df_grouped = df.groupby(['Week/Year', 'ID']).MDY.nunique()
Week/Year ID MDY
6/2019 15.0 5
The end result I'm looking for is to merge the MDY sums back to the initial dataframe by joining on Week and ID - I tried a few different ways: 我想要的最终结果是通过加入Week和ID将MDY总和合并回初始数据帧-我尝试了几种不同的方法:
input1.merge(df_grouped.to_frame(), left_on=['ID','Week/Year'], right_index=True)
to get something like: 得到类似的东西:
ID Datetime Week/Year MDY
0 15.0 2019-02-04 08:28:44 6/2019 5
1 15.0 2019-02-04 12:48:05 6/2019 5
2 15.0 2019-02-04 12:54:29 6/2019 5
3 15.0 2019-02-05 08:05:51 6/2019 5
4 15.0 2019-02-05 12:47:26 6/2019 5
5 15.0 2019-02-05 14:45:34 6/2019 5
After the join I just end up getting NaN across the board. 加入之后,我最终得到了NaN的全面支持。 Anyone able to steer me in the right direction?
有人能够引导我朝正确的方向前进吗?
Thanks. 谢谢。
this groupby
这个
groupby
df_grouped = df.groupby(['Week/Year', 'WD: Employee ID']).MDY.nunique()
should return a series has index as Week/Year WD: Employee ID
应该返回一个序列,其索引为
Week/Year WD: Employee ID
Week/Year WD: Employee ID
6/2019 15.0 5
Name: MDY , dtype: int64
However, you show its index as Week/Year ID
. 但是,您将其索引显示为
Week/Year ID
。 You may check columns name to make sure it match. 您可以检查列名以确保其匹配。
Next, on this 接下来,在此
input1.merge(df_grouped.to_frame(), left_on=['ID','Week/Year'], right_index=True)
Assume df_grouped
has index as you show in example which is Week/Year ID
, you have left_on
wrong order against right_index
. 假设
df_grouped
索引如示例中所示,即Week/Year ID
,则对right_index
left_on
顺序错误。 It should be 它应该是
input1.merge(df_grouped.to_frame(), left_on=['Week/Year', 'ID'], right_index=True)
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.