我试图将每月的数据点集转换为每周的数据点集,但为了实现这一目标,我将数据集分解为每日数据,然后将其聚合到周级别。当聚合发生时(通过groupby),我无法将数据分解为每日级别。

Month_End_Date  A   B   C   D
2/28/2019   Pikachu Starter 100000  5302
2/28/2019   Jolteon Evolution   250000  7935
3/31/2019   Charmander  Starter 62810   5103
3/31/2019   Bulbasaur   Starter 16868   6035
4/30/2019   Flareon Evolution   62810   5103
4/30/2019   Eevee   Starter 16868   6035
5/31/2019   Glaceon Evolution   62810   5103
5/31/2019   Leafeon Evolution   16868   6035
6/30/2019   Umbreon Evolution   62810   5103
6/30/2019   Espeon  Evolution   16868   6035

我想把第一行转换成
Month_End_Date  A   B   C   D
2/1/2019    Pikachu Starter 3571.428571 189.3571429
2/2/2019    Pikachu Starter 3571.428571 189.3571429
2/3/2019    Pikachu Starter 3571.428571 189.3571429
2/4/2019    Pikachu Starter 3571.428571 189.3571429
2/5/2019    Pikachu Starter 3571.428571 189.3571429

其中,日值除以28(因为二月有28天)
我已经搜索了ffill,但无法完全解决问题

最佳答案

首先通过Month_End_Date删除每列DataFrame.drop_duplicates中的重复项,然后通过向前填充缺失值和每月和每年仅最后一个筛选DataFrame.resample行来删除28

#convert column to datetimes and then to first day of month
df['Month_End_Date'] = (pd.to_datetime(df['Month_End_Date'], format='%m/%d/%Y')
                         .dt.to_period('m').dt.to_timestamp())
df = df.drop_duplicates('Month_End_Date').set_index('Month_End_Date')
#for duplicated last row of data
df.loc[df.index[-1] + pd.offsets.MonthEnd(1)] = df.iloc[-1]
df = df.resample('d').ffill()

df1 = df[df.groupby(df.index.to_period('m')).cumcount() < 28]
print (df1.tail())
                      A          B      C     D
Month_End_Date
2019-06-24      Umbreon  Evolution  62810  5103
2019-06-25      Umbreon  Evolution  62810  5103
2019-06-26      Umbreon  Evolution  62810  5103
2019-06-27      Umbreon  Evolution  62810  5103
2019-06-28      Umbreon  Evolution  62810  5103

如果需要所有值,不仅每个组首先使用GroupBy.cumcountresamplechain withgroupby计数器创建助手列:
df['Month_End_Date'] = (pd.to_datetime(df['Month_End_Date'], format='%m/%d/%Y')
                         .dt.to_period('m').dt.to_timestamp())
df['g'] = df.groupby('Month_End_Date').cumcount()
df = df.set_index('Month_End_Date')
df.loc[df.index[-1] + pd.offsets.MonthEnd(1)] = df.iloc[-1]

df = df.groupby('g').resample('d').ffill().reset_index(level=0, drop=True)
df2 = df[df.groupby(['g', df.index.to_period('m')]).cumcount() < 28]
print (df2.tail())
                     A          B      C     D  g
Month_End_Date
2019-06-24      Espeon  Evolution  16868  6035  1
2019-06-25      Espeon  Evolution  16868  6035  1
2019-06-26      Espeon  Evolution  16868  6035  1
2019-06-27      Espeon  Evolution  16868  6035  1
2019-06-28      Espeon  Evolution  16868  6035  1

关于python - 在Python中将每月数据转换为每日数据,我们在Stack Overflow上找到一个类似的问题:https://stackoverflow.com/questions/56731914/

10-09 17:07
查看更多