Question

我有以下df：

user_id    step    date
1         start    2018-04-17 15:27:07
1         step1    2018-04-17 15:28:07
1         end      2018-04-17 15:29:07
2         start    2018-05-17 15:28:07
2         step1    2018-05-17 15:29:07
2         end      2018-05-17 15:30:07

，我需要将其转换为下表：

user_id   start                  end                   time (end-start)
1         2018-04-17 15:27:07    2018-04-17 15:29:07   2
2         2018-05-17 15:28:07    2018-05-17 15:30:07   2

我被困在这一方面，任何帮助将不胜感激。

Answer 1

您可以旋转并找到timedelta

new_df = df.pivot('user_id', 'step', 'date').drop('step1', 1).reset_index()
new_df.columns.name = None
new_df['time (end-start)'] = (new_df['end'] - new_df['start']).astype('timedelta64[m]')


    user_id end                 start               time (end-start)
0   1       2018-04-17 15:29:07 2018-04-17 15:27:07 2.0
1   2       2018-05-17 15:30:07 2018-05-17 15:28:07 2.0

编辑：对于具有重复条目的数据框，如下所示：

    user_id step    date
0   1   start   2018-04-17 15:27:07
1   1   step1   2018-04-17 15:28:07
2   1   end     2018-04-17 15:29:07
3   1   end     2018-04-17 15:32:07
4   2   start   2018-05-17 15:26:07
5   2   start   2018-05-17 15:28:07
6   2   step1   2018-05-17 15:29:07
7   2   end     2018-05-17 15:30:07

new_df = df.pivot_table(index = 'user_id', columns = 'step', values = 'date', aggfunc = 'first').drop('step1', 1).reset_index() 

new_df.columns.name = None

new_df['time (end-start)'] = (new_df['end'] - new_df['start']).astype('timedelta64[m]')

你得到

    user_id end                 start               time (end-start)
0   1       2018-04-17 15:29:07 2018-04-17 15:27:07 2.0
1   2       2018-05-17 15:30:07 2018-05-17 15:26:07 4.0

熊猫枢轴日期

1 个答案: