问题
I have a csv file consisting of last 3 years of timeseries monthly data. Based on today's date, I would like to read only the previous 2 years of data for forecasting the future.
Data file example (has data from 01-01-15 to 31-10-19):
Date,Value
01-01-17,2
01-02-17,5
01-03-17,8
01-04-17,4
01-05-17,2
01-06-17,9
01-07-17,8
01-08-17,7
01-09-17,5
01-10-17,1
01-11-17,2
01-12-17,3
01-01-18,5
01-02-18,6
01-03-18,8
01-04-18,2
01-05-18,5
01-06-18,6
Desired result:
If today's date is 01/01/19, I want my training data to be data from 01/01/17 - 31/12/18.
I tried:df[df['date'] > (pd.to_datetime('2019-01-01', format = '%Y-%m-%d') - relativedelta(years = 2))]
However, I am getting data from 01-01-17 to 31-10-19(last record) instead of from 01-01-17 to 31-12-18.
回答1:
You can try the following
>>> from dateutil.relativedelta import relativedelta
>>> df[df.Date > datetime.now() - relativedelta(years=2)]
Date Value
12 2018-01-01 5
13 2018-01-02 6
14 2018-01-03 8
15 2018-01-04 2
16 2018-01-05 5
17 2018-01-06 6
Update
>>> from dateutil.relativedelta import relativedelta
>>> from datetime import date
>>> start_date = pd.Timestamp(datetime.now() - relativedelta(years=2))
>>> end_date = pd.Timestamp(date(date.today().year-1, 12, 31))
>>> df[(df.Date >= start_date) & (df.Date <= end_date)]
Date Value
12 2018-01-01 5
13 2018-01-02 6
14 2018-01-03 8
15 2018-01-04 2
16 2018-01-05 5
17 2018-01-06 6
来源:https://stackoverflow.com/questions/59491093/how-to-extract-data-from-previous-2-years-based-on-particular-date-in-python