Question

我在不同的时间间隔获得了几个Pandas DataFrame。一个是每日水平：

DatetimeIndex(['2007-12-01', '2007-12-02', '2007-12-03', '2007-12-04',
               '2007-12-05', '2007-12-06', '2007-12-07', '2007-12-08',
               '2007-12-09', '2007-12-10',
               ...
               '2016-08-22', '2016-08-23', '2016-08-24', '2016-08-25',
               '2016-08-26', '2016-08-27', '2016-08-28', '2016-08-29',
               '2016-08-30', '2016-08-31'],
              dtype='datetime64[ns]', length=3197, freq=None)

其他人处于非日常水平（他们总是不如每天坚决）。例如，这是每周一次：

DatetimeIndex(['2007-01-01', '2007-01-08', '2007-01-15', '2007-01-22',
               '2007-01-29', '2007-02-05', '2007-02-12', '2007-02-19',
               '2007-02-26', '2007-03-05',
               ...
               '2010-03-08', '2010-03-15', '2010-03-22', '2010-03-29',
               '2010-04-05', '2010-04-12', '2010-04-19', '2010-04-26',
               '2010-05-03',        'NaT'],
              dtype='datetime64[ns]', name='week', length=176, freq=None)

这是每月一次：

DatetimeIndex(['2013-04-01', '2013-05-01', '2013-06-01', '2013-07-01',
               '2013-08-01', '2013-09-01', '2013-10-01', '2013-11-01',
               '2013-12-01', '2014-01-01', '2014-02-01', '2014-03-01',
               '2014-04-01', '2014-05-01', '2014-06-01', '2014-07-01',
               '2014-08-01', '2014-09-01', '2014-10-01', '2014-11-01',
               '2014-12-01', '2015-01-01', '2015-02-01', '2015-03-01',
               '2015-04-01', '2015-05-01', '2015-06-01', '2015-07-01',
               '2015-08-01', '2015-09-01', '2015-10-01', '2015-11-01',
               '2015-12-01', '2016-01-01', '2016-02-01', '2016-03-01',
               '2016-04-01', '2016-05-01', '2016-06-01', '2016-07-01',
               '2016-08-01'],
              dtype='datetime64[ns]', name='month', freq=None)

这只是一个不规则间隔的奇怪球：

DatetimeIndex(['2014-02-14', '2014-05-08', '2014-09-19', '2014-09-24',
               '2015-01-21', '2016-05-26', '2016-06-02', '2016-06-04'],
              dtype='datetime64[ns]', name='date', freq=None)

我需要做的是将每日数据重新采样（求和）到其他人指定的时间间隔。因此，如果DatetimeIndex是每月一次，我需要将每日数据重新采样为每月。如果是每周一次，则应每周重新采样一次。如果它不规则，则需要匹配。我需要这个，因为我建立了这些数据的统计模型，我需要基本事实与观察值对齐。

如何让Pandas重新取样DataFrame df1，以匹配另一个任意数据框df2的DatetimeIndex？我一直在搜索，但我无法解决这个问题。它似乎是一个非常常见的熊猫任务，所以我必须错过一些东西。谢谢！

Answer 1

考虑使用pandas DataFrame.resample()：

# EXAMPLE DATA OF SEQUENTIAL DATES AND RANDOM NUMBERS
index = pd.date_range('12/01/2007', periods=3197, freq='D', dtype='datetime64[ns]')
series = pd.Series(np.random.randint(0,100, 3197), index=index)
df = pd.DataFrame({'num':series})
#             num
# 2007-12-01   73
# 2007-12-02   17
# 2007-12-03   63
# 2007-12-04   72
# 2007-12-05    4
# 2007-12-06   91
# 2007-12-07   20
# 2007-12-08   99
# 2007-12-09   97
# 2007-12-10   33

wdf = df.resample('W-SAT').sum()        # SATURDAY WEEK START
#             num
# 2007-12-01   73
# 2007-12-08  366
# 2007-12-15  354
# 2007-12-22  302
# 2007-12-29  310
# 2008-01-05  323
# 2008-01-12  424

mdf = df.resample('MS').sum()           # MONTH START
#              num
# 2007-12-01  1568
# 2008-01-01  1465
# 2008-02-01  1317
# 2008-03-01  1473
# 2008-04-01  1762
# 2008-05-01  1698
# 2008-06-01  1345

对于不规则间隔，请使用DataFrame.apply()中的自定义函数创建 enddate 列，该列将是当前行的日期间隔的结束截止日期串联（即 2015-01-01 的结束日期为 2015-01-21 在Datetimeindex系列中），使用系列过滤器计算。然后，在新的 enddate 列上运行groupby()进行汇总聚合：

irrdt = pd.DatetimeIndex(['2014-02-14', '2014-05-08', '2014-09-19', '2014-09-24',
                          '2015-01-21', '2016-05-26', '2016-06-02', '2016-06-04'],
                           dtype='datetime64[ns]', name='date', freq=None)    
def findrng(row):                      
    ed = str(irrdt[irrdt > row['Date']].min())[0:10]
    row['enddt'] = ed if ed !='NaT' else str(irrdt.max())[0:10]
    return(row)

df['Date'] = df.index
df = df.apply(findrng, axis=1).groupby(['enddt']).sum()    
#                num
# enddt             
# 2014-02-14  112143
# 2014-05-08    3704
# 2014-09-19    5958
# 2014-09-24     365
# 2015-01-21    5730
# 2016-05-26   24126
# 2016-06-02     305
# 2016-06-04    4142

如何重新取样DataFrame以使其与其他DataFrame正确对齐？

1 个答案: