Cfrgtkky

I need to clean a dataset that is similar to the following:

https://i.stack.imgur.com/yMjuy.png

The expected result looks like this:

https://i.stack.imgur.com/QFJHS.png

In other words, I need to keep all rows that (1) have 'action' in the column 'ACTION' and the immediate next one - (2) if the next one has 'result' in that column.
I've tried several combination of .shift(), but it did not work.

Thanks in advance.

You could try something like this:

df.assign(step=df['Action'].eq('action').cumsum())

  .drop_duplicates(subset=['Action','step'], keep='first')

  .groupby('step')

  .filter(lambda x: x.step.count()==2)

Output:

    SESSION  Action  TIME  step

0         1  action   0.1     1

1         1  result   0.2     1

2         2  action   0.1     2

3         2  result   0.2     2

4         3  action   0.3     3

5         3  result   0.4     3

7         1  action   0.3     4

8         1  result   0.4     4

12        3  action   0.6     6

13        3  result   0.7     6

14        4  action   0.8     7

15        4  result   0.9     7