熊猫填补时间序列中缺失的日期

Question

我有一个数据框，其中汇总了几天的数据。我想补充缺失的日子

我正在关注另一个帖子 Add missing dates to pandas dataframe，不幸的是，它覆盖了我的结果（可能功能略有改变？）...代码如下

import random
import datetime as dt
import numpy as np
import pandas as pd

def generate_row(year, month, day):
    while True:
        date = dt.datetime(year=year, month=month, day=day)
        data = np.random.random(size=4)
        yield [date] + list(data)

# days I have data for
dates = [(2000, 1, 1), (2000, 1, 2), (2000, 2, 4)]
generators = [generate_row(*date) for date in dates]

# get 5 data points for each
data = [next(generator) for generator in generators for _ in range(5)]

df = pd.DataFrame(data, columns=['date'] + ['f'+str(i) for i in range(1,5)])

# df
groupby_day = df.groupby(pd.PeriodIndex(data=df.date, freq='D'))
results = groupby_day.sum()

idx = pd.date_range(min(df.date), max(df.date))
results.reindex(idx, fill_value=0)

填写缺失日期索引前的结果

之后的结果

Answer 1

您需要使用period_range 而不是date_range：

In [11]: idx = pd.period_range(min(df.date), max(df.date))
    ...: results.reindex(idx, fill_value=0)
    ...:
Out[11]:
                  f1        f2        f3        f4
2000-01-01  2.049157  1.962635  2.756154  2.224751
2000-01-02  2.675899  2.587217  1.540823  1.606150
2000-01-03  0.000000  0.000000  0.000000  0.000000
2000-01-04  0.000000  0.000000  0.000000  0.000000
2000-01-05  0.000000  0.000000  0.000000  0.000000
2000-01-06  0.000000  0.000000  0.000000  0.000000
2000-01-07  0.000000  0.000000  0.000000  0.000000
2000-01-08  0.000000  0.000000  0.000000  0.000000
2000-01-09  0.000000  0.000000  0.000000  0.000000
2000-01-10  0.000000  0.000000  0.000000  0.000000
2000-01-11  0.000000  0.000000  0.000000  0.000000
2000-01-12  0.000000  0.000000  0.000000  0.000000
2000-01-13  0.000000  0.000000  0.000000  0.000000
2000-01-14  0.000000  0.000000  0.000000  0.000000
2000-01-15  0.000000  0.000000  0.000000  0.000000
2000-01-16  0.000000  0.000000  0.000000  0.000000
2000-01-17  0.000000  0.000000  0.000000  0.000000
2000-01-18  0.000000  0.000000  0.000000  0.000000
2000-01-19  0.000000  0.000000  0.000000  0.000000
2000-01-20  0.000000  0.000000  0.000000  0.000000
2000-01-21  0.000000  0.000000  0.000000  0.000000
2000-01-22  0.000000  0.000000  0.000000  0.000000
2000-01-23  0.000000  0.000000  0.000000  0.000000
2000-01-24  0.000000  0.000000  0.000000  0.000000
2000-01-25  0.000000  0.000000  0.000000  0.000000
2000-01-26  0.000000  0.000000  0.000000  0.000000
2000-01-27  0.000000  0.000000  0.000000  0.000000
2000-01-28  0.000000  0.000000  0.000000  0.000000
2000-01-29  0.000000  0.000000  0.000000  0.000000
2000-01-30  0.000000  0.000000  0.000000  0.000000
2000-01-31  0.000000  0.000000  0.000000  0.000000
2000-02-01  0.000000  0.000000  0.000000  0.000000
2000-02-02  0.000000  0.000000  0.000000  0.000000
2000-02-03  0.000000  0.000000  0.000000  0.000000
2000-02-04  1.856158  2.892620  2.986166  2.793448

这是因为您的 groupby 使用的是 PeriodIndex，而不是 datetime：

df.groupby(pd.PeriodIndex(data=df.date, freq='D'))

您可以改用pd.Grouper:

df.groupby(pd.Grouper(key="date", freq='D'))

这会给出一个日期时间索引。

Answer 2

来自cᴏʟᴅsᴘᴇᴇᴅ在cmets中的提示：

---

resample 很适合这里。

Resample：时间序列变频重采样的便捷方法。对象必须具有类似日期时间的索引（DatetimeIndex、PeriodIndex 或 TimedeltaIndex），或者将类似日期时间的值传递给 on 或 level 关键字。

import random
import datetime as dt
import numpy as np
import pandas as pd

def generate_row(year, month, day):
    while True:
        date = dt.datetime(year=year, month=month, day=day)
        data = np.random.random(size=4)
        yield [date] + list(data)

# days I have data for
dates = [(2000, 1, 1), (2000, 1, 2), (2000, 2, 4)]
generators = [generate_row(*date) for date in dates]

# get 5 points for each
data = [next(generator) for generator in generators for _ in range(5)]

# make dataframe
df = pd.DataFrame(data, columns=['date'] + ['f'+str(i) for i in range(1,5)])

# using the resample method
df.set_index(df.date, inplace=True)
df = df.resample('D').sum().fillna(0)