Python：计算一段时间内 Pandas 数据框中的累积量

Question

目标：计算自 2020-01-01 以来的累计收入。

我有一个 python 字典，如下所示

data = [{"game_id":"Racing","user_id":"ABC123","amt":5,"date":"2020-01-01"},

    {"game_id":"Racing","user_id":"ABC123","amt":1,"date":"2020-01-04"},

    {"game_id":"Racing","user_id":"CDE123","amt":1,"date":"2020-01-04"},

    {"game_id":"DH","user_id":"CDE123","amt":100,"date":"2020-01-03"},

    {"game_id":"DH","user_id":"CDE456","amt":10,"date":"2020-01-02"},

    {"game_id":"DH","user_id":"CDE789","amt":5,"date":"2020-01-02"},

    {"game_id":"DH","user_id":"CDE456","amt":1,"date":"2020-01-03"},

    {"game_id":"DH","user_id":"CDE456","amt":1,"date":"2020-01-03"}]

上面的同一个字典看起来像一个表

   game_id   user_id  amt  activity date

  'Racing', 'ABC123', 5,   '2020-01-01'

  'Racing', 'ABC123', 1,   '2020-01-04'

  'Racing', 'CDE123', 1,   '2020-01-04'

  'DH',     'CDE123', 100, '2020-01-03'

  'DH',     'CDE456', 10,  '2020-01-02'

  'DH', '    CDE789', 5,   '2020-01-02'

  'DH',     'CDE456', 1,   '2020-01-03'

  'DH',     'CDE456', 1,   '2020-01-03'

年龄计算为交易日期与 2020-01-01 之间的差异。付款人总数是每场比赛的付款人数量。

我正在尝试创建一个数据框，其中包含从第一笔交易之日到交易第二天的每一天的累积结果。例如：对于 game_id Racing，我们在 2020 年 1 月 1 日从金额 5 开始，所以年龄为 0。在 2020 年 1 月 2 日，金额仍然是 5，因为我们那天没有交易。在 2020 年 1 月 3 日，金额为 5。但在 2020 年 1 月 4 日，金额为 7，因为我们在这一天有 2 笔交易。

预期产出

Game    Age    Cum_rev    Total_unique_payers_per_game

Racing  0      5          2

Racing  1      5          2

Racing  2      5          2

Racing  3      7          2

DH      0      0          3

DH      1      15         3

DH      2      117        3

DH      3      117        3

如何在 python 中使用窗口函数，就像我们在 SQL 中使用一样。有没有更好的方法来解决这个问题？

Answer 1

这里非常复杂的部分是填写日期。我使用了申请，但我不确定这是最好的方法import pandas as pddata = [{"game_id":"Racing","user_id":"ABC123","amt":5,"date":"2020-01-01"},        {"game_id":"Racing","user_id":"ABC123","amt":1,"date":"2020-01-04"},        {"game_id":"Racing","user_id":"CDE123","amt":1,"date":"2020-01-04"},        {"game_id":"DH","user_id":"CDE123","amt":100,"date":"2020-01-03"},        {"game_id":"DH","user_id":"CDE456","amt":10,"date":"2020-01-02"},        {"game_id":"DH","user_id":"CDE789","amt":5,"date":"2020-01-02"},        {"game_id":"DH","user_id":"CDE456","amt":1,"date":"2020-01-03"},        {"game_id":"DH","user_id":"CDE456","amt":1,"date":"2020-01-03"}]df = pd.DataFrame(data)# we want datetime not objectdf["date"] = df["date"].astype("M8[us]")# we will need to merge this at the endgrp = df.groupby("game_id")['user_id']\        .nunique()\        .reset_index(name="Total_unique_payers_per_game")# sum amt per game_id datedf = df.groupby(["game_id", "date"])["amt"].sum().reset_index()# dates from 2020-01-01 till the max date in dfdates = pd.DataFrame({"date": pd.date_range("2020-01-01", df["date"].max())})# add missing datesdef expand_dates(x):    x = pd.merge(dates, x.drop("game_id", axis=1), how="left")    x["amt"] = x["amt"].fillna(0)    return xdf = df.groupby("game_id")\       .apply(expand_dates)\       .reset_index().drop("level_1", axis=1)df["Cum_rev"] = df.groupby("game_id")['amt'].transform("cumsum")# this is equivalent as long as data is sorted# df["Cum_rev"] = df.groupby("game_id")['amt'].cumsum()# merge unique payers per gamedf = pd.merge(df, grp, how="left")# dates differencedf["Age"] = "2020-01-01"df["Age"] = df["Age"].astype("M8[us]")df["Age"] = (df["date"]-df["Age"]).dt.days# then you can eventually filterdf = df[["game_id", "Age",          "Cum_rev", "Total_unique_payers_per_game"]]\       .rename(columns={"game_id":"Game"})