Remove lags/gaps in time series data for a particular dataframe in pandas

user_16237276 0
Viewed 126

I am trying to model a distribution for a particular stock in terms of the amount of times the order book is updated at a certain timeframe.

The issue I am having is in relation to data engineering and pandas. Because I am working only with trading hours and trading days, my dataset has multiple gaps and so the data does not appear continuous. The graph below clearly shows that:

enter image description here

You can see that the large gaps are weekends and the small gaps are post trading hours. The small black squares (if zoomed in) would look something like this: enter image description here

The dataframe would look like this:

    arrivalTime           value           date
    0 days 09:30:02.231     1          2021-05-03
    0 days 09:30:02.981     3          2021-05-03
    0 days 09:30:02.999     99         2021-05-03
    0 days 09:30:10.284     11         2021-05-03
    0 days 09:30:10.293     92         2021-05-03
... ... ...
    0 days 15:59:42.654     82         2021-05-28
    0 days 15:59:42.655     19         2021-05-28
    0 days 15:59:42.651     122        2021-05-28
    0 days 15:59:42.941     199        2021-05-28
    0 days 15:59:44.721     19         2021-05-28

The exact thing that I would need is that once a trading day is over, the next day continues exactly after that day has ended. Let me know if there are any questions

Thanks!

pandas time-series dataframe statistics python
1 Answers

IIUC, you want to shift dates to have continuous dates:

Sample:

>>> df
         date
0  2021-05-05
1  2021-05-05
2  2021-05-05
3  2021-05-06
4  2021-05-06
5  2021-05-06
6  2021-05-07
7  2021-05-07
8  2021-05-07
9  2021-05-10  # <- 2021-05-08
10 2021-05-10  # <- 2021-05-08
11 2021-05-10  # <- 2021-05-08
12 2021-05-11  # <- 2021-05-09
13 2021-05-11  # <- 2021-05-09
14 2021-05-11  # <- 2021-05-09

>>> df['date'].min() + df['date'].diff().ne(pd.Timedelta(0)).cumsum().sub(1) \
                                 .apply(pd.tseries.offsets.Day)
0    2021-05-05
1    2021-05-05
2    2021-05-05
3    2021-05-06
4    2021-05-06
5    2021-05-06
6    2021-05-07
7    2021-05-07
8    2021-05-07
9    2021-05-08
10   2021-05-08
11   2021-05-08
12   2021-05-09
13   2021-05-09
14   2021-05-09
Name: date, dtype: datetime64[ns]
user_152399510
Related