使用另一个数据框中的值和日期范围创建全新的数据框

Posted 2023-02-23

【中文标题】使用另一个数据框中的值和日期范围创建全新的数据框

【英文标题】：Create brand new dataframe using values and date ranges within another dataframe

【发布时间】：2021-10-26 22:24:15

【问题描述】：

我有一个由经常性费用组成的数据框，包括开始日期 (purchase_date)、结束日期 (date_terminated)、ID 和价格：

       id    price purchase_date date_terminated
0   AA11      100    2019-03-29             NaT
1   AA12  10750.0    2020-02-28             NaT
2   AA13   2500.0    2020-06-01             NaT
3   BB11    600.0    2020-06-01      2021-08-01
4   BB12    600.0    2020-06-01      2021-06-17
5   BB13   6692.0    2020-07-08      2021-04-01
6   CC11   6692.0    2020-08-12             NaT
7   CC12   6692.0    2020-08-12             NaT
8   CC13    600.0    2020-09-01      2021-04-01
9   DD11    600.0    2020-09-01             NaT

如果 date_terminated==NaT，则表示费用仍在重复发生。

我还有一个日期列表，从我最早的经常性费用的开始日期开始，一直到用户选择的任何日期：

[datetime.datetime(2019, 3, 15, 0, 0),
 datetime.datetime(2019, 4, 15, 0, 0),
                  .
                  .
                  . 
                  .
                  .
 datetime.datetime(2021, 6, 15, 0, 0),
 datetime.datetime(2021, 7, 15, 0, 0),
 datetime.datetime(2021, 8, 15, 0, 0)]

我想构建一个数据框，其中 index 作为 dates_list，列作为费用 ID，并使用 purchase_date 和 date_terminated 作为参考点在整个 df 中分配我的费用。

最终结果应该类似于以下内容：

             AA11     AA12  AA13  BB11 BB12 BB13 CC11 CC12 CC13 DD11
2019-03-15    NaN      NaN   NaN   NaN  NaN  NaN  NaN  NaN  NaN  NaN 
2019-04-15    100      NaN   NaN   NaN  NaN  NaN  NaN  NaN  NaN  NaN
2019-05-15    100      NaN   NaN   NaN  NaN  NaN  NaN  NaN  NaN  NaN
                               .
                               .
                               .
                               .     
2021-06-15    100    10750  2500   600  600  NaN  6692 6692 NaN  600
2021-07-15    100    10750  2500   600  NaN  NaN  6692 6692 NaN  600
2021-08-15    100    10750  2500   NaN  NaN  NaN  6692 6692 NaN  600

【参考方案1】：

试试：

import datetime

df["purchase_date"] = pd.to_datetime(df["purchase_date"])
df["date_terminated"] = pd.to_datetime(df["date_terminated"])

lst = [
    datetime.datetime(2019, 3, 15, 0, 0),
    datetime.datetime(2019, 4, 15, 0, 0),
    datetime.datetime(2021, 6, 15, 0, 0),
    datetime.datetime(2021, 7, 15, 0, 0),
    datetime.datetime(2021, 8, 15, 0, 0),
]

df_tmp = pd.DataFrame("dates": lst)

mx = df_tmp.dates.max()
df["purchase_date"] = df.apply(
    lambda x: pd.date_range(
        x["purchase_date"],
        x["date_terminated"] if pd.notna(x["date_terminated"]) else mx,
    ),
    axis=1,
)
cols = df.id.unique()
df = df.explode("purchase_date")
df = df[df.purchase_date.isin(df_tmp.dates)]
print(
    df.pivot(index="purchase_date", columns="id", values="price")
    .reindex(df_tmp["dates"])
    .reindex(cols, axis=1)
)

打印：

id           AA11     AA12    AA13   BB11   BB12  BB13    CC11    CC12  CC13   DD11
dates                                                                              
2019-03-15    NaN      NaN     NaN    NaN    NaN   NaN     NaN     NaN   NaN    NaN
2019-04-15  100.0      NaN     NaN    NaN    NaN   NaN     NaN     NaN   NaN    NaN
2021-06-15  100.0  10750.0  2500.0  600.0  600.0   NaN  6692.0  6692.0   NaN  600.0
2021-07-15  100.0  10750.0  2500.0  600.0    NaN   NaN  6692.0  6692.0   NaN  600.0
2021-08-15  100.0  10750.0  2500.0    NaN    NaN   NaN  6692.0  6692.0   NaN  600.0