I have a datetime based dataframe as below,
timestamp value ... metric
36 2014-04-02 17:20:00 125.098263 ... 25.098263
14 2014-04-06 16:25:00 140.072787 ... 265.171050
10 2014-04-11 09:00:00 127.882020 ... 393.053070
45 2014-04-11 09:05:00 115.705719 ... 508.758789
24 2014-04-11 09:15:00 127.261178 ... 636.019967
17 2014-04-11 09:20:00 121.157997 ... 757.177965
49 2014-04-11 09:25:00 120.468468 ... 877.646433
8 2014-04-11 09:45:00 135.642696 ... 1013.289128
33 2014-04-11 09:55:00 125.210049 ... 1138.499178
19 2014-04-11 10:05:00 159.259713 ... 1297.758890
52 2014-04-11 10:20:00 150.082482 ... 1447.841373
I want to create new column named as 'diff_col' contains either 'same' or 'diff' values. If a date is not continuous, it will taken as 'diff' otherwise it is 'same'. In the above dataframe, 2014-04-02 17:20:00 and 2014-04-06 16:25:00 are different dates compare to remaining datetime values.
How to create the diff_col .
I tried, df['diff_col']=df.groupby(pd.Grouper(key = 'timestamp', freq='1D'))
but it didn't correctly create the expected column. My required dataframe is as below,
timestamp value ... metric diff_col
36 2014-04-02 17:20:00 125.098263 ... 25.098263 diff
14 2014-04-06 16:25:00 140.072787 ... 265.171050 diff
10 2014-04-11 09:00:00 127.882020 ... 393.053070 same
45 2014-04-11 09:05:00 115.705719 ... 508.758789 same
24 2014-04-11 09:15:00 127.261178 ... 636.019967 same
17 2014-04-11 09:20:00 121.157997 ... 757.177965 same
49 2014-04-11 09:25:00 120.468468 ... 877.646433 same
8 2014-04-11 09:45:00 135.642696 ... 1013.289128 same
33 2014-04-11 09:55:00 125.210049 ... 1138.499178 same
19 2014-04-11 10:05:00 159.259713 ... 1297.758890 same
52 2014-04-11 10:20:00 150.082482 ... 1447.841373 same
Please provide suggestion on this.
Thanks, Kumar
CodePudding user response:
You can compare the successive rows to see if this is the same date (extracted with dt.normalize
) and use this as grouper to get the size with groupby.transform('size')
, if the size is > 1, set 'same' else 'diff' with help of numpy.where
:
import numpy as np
# ensure datetime
df['timestamp'] = pd.to_datetime(df['timestamp'])
# get day
s = df['timestamp'].dt.normalize()
# compare successive rows and identify group size
df['diff_col'] = np.where(df.groupby(s.ne(s.shift()).cumsum())
.transform('size').gt(1),
'same', 'diff')
Output:
timestamp value ... metric diff_col
36 2014-04-02 17:20:00 125.098263 ... 25.098263 diff
14 2014-04-06 16:25:00 140.072787 ... 265.171050 diff
10 2014-04-11 09:00:00 127.882020 ... 393.053070 same
45 2014-04-11 09:05:00 115.705719 ... 508.758789 same
24 2014-04-11 09:15:00 127.261178 ... 636.019967 same
17 2014-04-11 09:20:00 121.157997 ... 757.177965 same
49 2014-04-11 09:25:00 120.468468 ... 877.646433 same
8 2014-04-11 09:45:00 135.642696 ... 1013.289128 same
33 2014-04-11 09:55:00 125.210049 ... 1138.499178 same
19 2014-04-11 10:05:00 159.259713 ... 1297.758890 same
52 2014-04-11 10:20:00 150.082482 ... 1447.841373 same