calculating over partition in pandas dataframe

Question

I have a table like so:

ID  Timestamp   Status
A   5/30/2022 2:29  Run Ended
A   5/30/2022 0:23  In Progress
A   5/30/2022 0:22  Prepared
B   5/30/2022 11:15 Run Ended
B   5/30/2022 9:18  In Progress
B   5/30/2022 0:55  Prepared

I want to compute the duration between each status grouped by the ID. So the resulting output table would be:

ID  Duration(min)   Status change
A   0.40    In Progress-Prepared
A   125.82  Run Ended - In Progress
B   502.78  In Progress-Prepared
B   117.34  Run Ended - In Progress

How do I order it by descending timestamp (grouped by ID) and then subtract the last row from the previous row all the way upto the top for each ID group?

How do you compute the values? A from prepared to in progress is -1 min. — keramat, May 31 '22 at 04:09
@keramat Sorry I meant it the other way. Its just the transition duration. — thentangler, May 31 '22 at 15:09

score 2 · Answer 1 · answered May 31 '22 at 04:32

You can use a groupby.diff and groupby.shift:

out = (df
 .assign(**{'Duration(min)': pd.to_datetime(df['Timestamp'], dayfirst=False)
            .groupby(df['ID'])
            .diff(-1).dt.total_seconds() # diff in seconds to next time in group
            .div(60), # convert to minutes
           'Status change': df.groupby('ID')['Status'].shift(-1)+'-'+df['Status']
           })
 .dropna(subset='Duration(min)') # get rid of empty rows
 [['ID', 'Duration(min)', 'Status change']]
 )

Output:

  ID  Duration(min)          Status change
0  A          126.0  In Progress-Run Ended
1  A            1.0   Prepared-In Progress
3  B          117.0  In Progress-Run Ended
4  B          503.0   Prepared-In Progress

score 1 · Accepted Answer · answered May 31 '22 at 04:09

You can use groupby('ID')[value].shift(1) to access the previous value in the same ID group.

import pandas as pd

df = pd.DataFrame({
    'ID': ['a','a','a','b','b','b'],
    'time': [1,2,3,1,4,5],
    'status': ['x','y','z','xx','yy','zz']
})
df['previous_time'] = df.groupby('ID')['time'].shift(1)
df['previous_status'] = df.groupby('ID')['status'].shift(1)
df = df.dropna()
df['duration'] = df['time'] - df['previous_time'] # change this line to calculate duration between time instead
df['status_change'] = df['previous_status'] + '-' + df['status']
print (df[['ID','duration','status_change']].to_markdown(index=False))

Output:

ID	duration	status_change
a	1	x-y
a	1	y-z
b	3	xx-yy
b	1	yy-zz

PS. you can subtract time and previous_time with answer in this thread

calculating over partition in pandas dataframe

2 Answers2