add grouped cumulative sum column to pandas dataframe

pandas-groupby-cumsum.py

"""
add grouped cumulative sum column to pandas dataframe

Add a new column to a pandas dataframe which holds the cumulative sum for a given grouped window

Desired output:
user_id,day,session_minutes,cumulative_minutes
516530,0,NaN,0
516530,1,0,0
516532,0,5,5
516532,1,33,38
516534,0,32,32
516534,1,104,136
516535,0,5,5
516535,1,0,5
"""
import pandas as pd
from StringIO import StringIO

data = """
user_id,day,session_minutes
516530,0,NaN
516530,1,0
516532,0,5
516532,1,33
516534,0,32
516534,1,104
516535,0,5
516535,1,0
"""

df = pd.read_csv(StringIO(data))
cumsums = df.groupby(['user_id', 'day']).sum().fillna(0).groupby(level=0).cumsum()
df.set_index(['user_id', 'day'], inplace=True)
df['cumulative_minutes'] = cumsums
df.reset_index(inplace=True)
print(df)

#   user_id  day  session_minutes  cumulative_minutes
# 0   516530    0              NaN                   0
# 1   516530    1                0                   0
# 2   516532    0                5                   5
# 3   516532    1               33                  38
# 4   516534    0               32                  32
# 5   516534    1              104                 136
# 6   516535    0                5                   5
# 7   516535    1                0                   5

Hello,

Thank you for sharing. Trying your example with

data="""
PATTERN_NAME | PROD_INJ_DATE | OIL_PROD
_CRCU_1101 | 1/31/1973 | 4752
_CRCU_1101 | 2/28/1973 | 4428
_CRCU_1101 | 3/31/1973 | 5316
_CRCU_1101 | 4/30/1973 | 4776
_CRCU_1101 | 5/31/1973 | 4860
_CRCU_1101 | 6/30/1973 | 4905
_CRCU_1101 | 7/31/1973 | 5208
_CRCU_1101 | 8/31/1973 | 5034
_CRCU_1101 | 9/30/1973 | 3240
_CRCU_1103 | 1/31/1973 | 372
_CRCU_1103 | 2/28/1973 | 345
_CRCU_1103 | 3/31/1973 | 397
_CRCU_1103 | 4/30/1973 | 373
_CRCU_1103 | 5/31/1973 | 432
_CRCU_1103 | 6/30/1973 | 454
_CRCU_1103 | 7/31/1973 | 483
_CRCU_1103 | 8/31/1973 | 467
_CRCU_1103 | 9/30/1973 | 458
_CRCU_1104 | 1/31/1973 | 447
_CRCU_1104 | 2/28/1973 | 414
_CRCU_1104 | 3/31/1973 | 488
_CRCU_1104 | 4/30/1973 | 449
_CRCU_1104 | 5/31/1973 | 432
_CRCU_1104 | 6/30/1973 | 454
_CRCU_1104 | 7/31/1973 | 482
_CRCU_1104 | 8/31/1973 | 467
_CRCU_1104 | 9/30/1973 | 431
"""

Using the code below

df=pd.read_csv("ProdByPattern.csv",delimiter="\t")
cumsums = df.groupby(['PATTERN_NAME','PROD_INJ_DATE']).sum().fillna(0).groupby(level=0).cumsum()
df.set_index(['PATTERN_NAME','PROD_INJ_DATE'], inplace=True)
df['cum_oil'] = cumsums
df.reset_index(inplace=True)
print(df)

I got the expected results.

PATTERN_NAME PROD_INJ_DATE OIL_PROD cum_oil
0 _CRCU_1101 1/31/1973 4752 4752
1 _CRCU_1101 2/28/1973 4428 9180
2 _CRCU_1101 3/31/1973 5316 14496
3 _CRCU_1101 4/30/1973 4776 19272
4 _CRCU_1101 5/31/1973 4860 24132
5 _CRCU_1101 6/30/1973 4905 29037
6 _CRCU_1101 7/31/1973 5208 34245
7 _CRCU_1101 8/31/1973 5034 39279
8 _CRCU_1101 9/30/1973 3240 42519
9 _CRCU_1103 1/31/1973 372 372
10 _CRCU_1103 2/28/1973 345 717
11 _CRCU_1103 3/31/1973 397 1114
12 _CRCU_1103 4/30/1973 373 1487
13 _CRCU_1103 5/31/1973 432 1919
14 _CRCU_1103 6/30/1973 454 2373
15 _CRCU_1103 7/31/1973 483 2856
16 _CRCU_1103 8/31/1973 467 3323
17 _CRCU_1103 9/30/1973 458 3781
18 _CRCU_1104 1/31/1973 447 447
19 _CRCU_1104 2/28/1973 414 861
20 _CRCU_1104 3/31/1973 488 1349
21 _CRCU_1104 4/30/1973 449 1798
22 _CRCU_1104 5/31/1973 432 2230
23 _CRCU_1104 6/30/1973 454 2684
24 _CRCU_1104 7/31/1973 482 3166
25 _CRCU_1104 8/31/1973 467 3633
26 _CRCU_1104 9/30/1973 431 4064