我如何跟踪帐户过去 6 个月收到的总交易金额？ [复制]答案

【问题标题】：How can I keep track of total transaction amount received by an account each last 6 month? [duplicate]我如何跟踪帐户过去 6 个月收到的总交易金额？ [复制]
【发布时间】：2020-12-21 17:36:08
【问题描述】：

这是我的交易数据。它显示了从from 列中的帐户到to 列中的帐户进行的交易以及日期和金额信息

data 

id          from    to          date        amount  
<int>       <fctr>  <fctr>      <date>      <dbl>
19521       6644    6934        2005-01-01  700.0
19524       6753    8456        2005-01-01  600.0
19523       9242    9333        2005-01-01  1000.0
…           …       …           …           …
1056317     7819    7454        2010-12-31  60.2
1056318     6164    7497        2010-12-31  107.5
1056319     7533    7492        2010-12-31  164.1

现在我想做如下：我想逐行检查每笔交易，对于from 列中的每个帐户，我想跟踪他们在过去 6 个月收到的交易金额进行特定交易的时间，并希望将此信息保存为新列。（因此，此新列将描述from 列中的帐户在交易日期前的最后六个月内收到的总交易金额。）

例如：
在第一行数据中，对于帐户6644，如果6644 在日期"2004-07-05"-"2005-01-01" 之间有一笔交易，我应该查看to 列，这是截至日期2005-01-01 的6 个月期间这是6644 进行交易的日期。如果有6644 收到的此类交易，我应该将它们汇总并将此总和信息添加到新列total_trx_amount_received_in_last_6month 作为值。同样，我应该为帐户 6753 做同样的事情并查找它在日期"2004-07-05"-"2005-01-01" 之间获得的交易并将它们相加并将此总和添加到total_trx_amount_received_in_last_6month 列中。而且，我应该以这种方式在数据中逐行继续。

那么，我怎样才能为整个数据实现这一点？

PS：在日期区间"2004-07-05"-"2005-01-01"，"2005-01-01"是交易日期，要得到第二个日期"2004-07-05"我从交易日期"2005-01-01"减去180天（约6个月）。

为了看得更清楚，我提供以下数据：
我还将展示输出将如何。假设我们只有这么多交易。这里只考虑5370 帐户，因为其他帐户8605,6390,8934 在这里没有收到任何交易。

id          from    to          date        amount  total_trx_amount_received_in_last_6month 
<int>       <fctr>  <fctr>      <date>      <dbl>    <dbl>
18529       5370    9356        2005-05-31  24.4     0.0
13742       5370    5605        2005-08-05  7618.0   0.0
9913        5370    8567        2005-09-12  21971.0  0.0
956         8605    5370        2005-10-05  5245.0   0.0
2557        5370    5636        2005-11-12  2921.0   5245.0    
1602        6390    5370        2005-11-26  8000.0   0.0
18669       5370    8933        2005-11-30  169.2    (5245.0+8000.0)=13245
35900       5370    8483        2006-01-31  71.5     (5245.0+8000.0)=13245
48667       8934    5370        2006-03-31  14.6     0.0
51341       5370    7626        2006-04-11  4214.0   (8000.0+14.6)=8014.6

这是我所做的：首先注意上面这个小数据是按照date升序排列的。
在第一行，对于from column 中的帐户5370，我查看过去的数据以查看5370 在日期"2004-12-02"-"2005-05-31" 之间是否收到任何交易。由于第一行是第一笔交易，显然5370 在日期"2005-05-31" 之前没有收到交易，所以我将0.0 分别记录到total_trx_amount_received_in_last_6month 列。在第二行中，对于from column 中的帐户5370，5370 在日期"2005-02-06"-"2005-08-05" 之间再次没有收到任何交易，因此我将0.0 登录到total_trx_amount_received_in_last_6month 列。同样，我分别在帐户 5370 和 8605 的第 3 行和第 4 行记录 0.0。在第五行，对于from column中的账户5370，5370在日期"2005-05-16"-"2005-11-12"之间收到一笔交易，在"2005-10-05"（数据第4行）收到一笔交易，金额为@ 987654368@ 所以我将5245.0 登录到total_trx_amount_received_in_last_6month 列。在第六行，对于from column 中的帐户6390，6390 在日期"2005-05-30"-"2005-11-26" 之间没有收到任何交易，所以我将0.0 登录到total_trx_amount_received_in_last_6month 列。数据的所有行都是这样的。

dput() 输出数据：

structure(list(id = c(18529L, 13742L, 9913L, 956L, 2557L, 1602L, 
18669L, 35900L, 48667L, 51341L, 53713L, 60126L, 60545L, 65113L, 
66783L, 83324L, 87614L, 88898L, 89874L, 94765L, 100277L, 101587L, 
103444L, 108414L, 113319L, 121516L, 126607L, 130170L, 131771L, 
135002L, 149431L, 157403L, 157645L, 158831L, 162597L, 162680L, 
163901L, 165044L, 167082L, 168562L, 168940L, 172578L, 173031L, 
173267L, 177507L, 179167L, 182612L, 183499L, 188171L, 189625L, 
193940L, 198764L, 199342L, 200134L, 203328L, 203763L, 204733L, 
205651L, 209672L, 210242L, 210979L, 214532L, 214741L, 215738L, 
216709L, 220828L, 222140L, 222905L, 226133L, 226527L, 227160L, 
228193L, 231782L, 232454L, 233774L, 237836L, 237837L, 238860L, 
240223L, 245032L, 246673L, 247561L, 251611L, 251696L, 252663L, 
254410L, 255126L, 255230L, 258484L, 258485L, 259309L, 259910L, 
260542L, 262091L, 264462L, 264887L, 264888L, 266125L, 268574L, 
272959L), from = c("5370", "5370", "5370", "8605", "5370", "6390", 
"5370", "5370", "8934", "5370", "5635", "6046", "5680", "8026", 
"9037", "5370", "7816", "8046", "5492", "8756", "5370", "9254", 
"5370", "5370", "7078", "6615", "5370", "9817", "8228", "8822", 
"5735", "7058", "5370", "8667", "9315", "6053", "7990", "8247", 
"8165", "5656", "9261", "5929", "8251", "5370", "6725", "5370", 
"6004", "7022", "7442", "5370", "8679", "6491", "7078", "5370", 
"5370", "5370", "5658", "5370", "9296", "8386", "5370", "5370", 
"5370", "9535", "5370", "7541", "5370", "9621", "5370", "7158", 
"8240", "5370", "5370", "8025", "5370", "5370", "5370", "6989", 
"5370", "7059", "5370", "5370", "5370", "9121", "5608", "5370", 
"5370", "7551", "5370", "5370", "5370", "5370", "9163", "9362", 
"6072", "5370", "5370", "5370", "5370", "5370"), to = c("9356", 
"5605", "8567", "5370", "5636", "5370", "8933", "8483", "5370", 
"7626", "5370", "5370", "5370", "5370", "5370", "9676", "5370", 
"5370", "5370", "5370", "9105", "5370", "9772", "6979", "5370", 
"5370", "7564", "5370", "5370", "5370", "5370", "5370", "8744", 
"5370", "5370", "5370", "5370", "5370", "5370", "5370", "5370", 
"5370", "5370", "7318", "5370", "8433", "5370", "5370", "5370", 
"7122", "5370", "5370", "5370", "8566", "6728", "9689", "5370", 
"8342", "5370", "5370", "5614", "5596", "5953", "5370", "7336", 
"5370", "7247", "5370", "7291", "5370", "5370", "6282", "7236", 
"5370", "8866", "8613", "9247", "5370", "6767", "5370", "9273", 
"7320", "9533", "5370", "5370", "8930", "9343", "5370", "9499", 
"7693", "7830", "5392", "5370", "5370", "5370", "7497", "8516", 
"9023", "7310", "8939"), date = structure(c(12934, 13000, 13038, 
13061, 13099, 13113, 13117, 13179, 13238, 13249, 13268, 13296, 
13299, 13309, 13314, 13391, 13400, 13404, 13409, 13428, 13452, 
13452, 13460, 13482, 13493, 13518, 13526, 13537, 13542, 13544, 
13596, 13616, 13617, 13626, 13633, 13633, 13639, 13642, 13646, 
13656, 13660, 13664, 13667, 13669, 13677, 13686, 13694, 13694, 
13707, 13716, 13725, 13738, 13739, 13746, 13756, 13756, 13756, 
13761, 13769, 13770, 13776, 13786, 13786, 13786, 13791, 13799, 
13806, 13813, 13817, 13817, 13817, 13822, 13829, 13830, 13836, 
13847, 13847, 13847, 13852, 13860, 13866, 13871, 13878, 13878, 
13878, 13882, 13883, 13883, 13887, 13887, 13888, 13889, 13890, 
13891, 13895, 13896, 13896, 13899, 13905, 13909), class = "Date"), 
    amount = c(24.4, 7618, 21971, 5245, 2921, 8000, 169.2, 71.5, 
    14.6, 4214, 14.6, 13920, 14.6, 24640, 1600, 261.1, 16400, 
    3500, 2700, 19882, 182, 14.6, 16927, 25653, 3059, 2880, 9658, 
    4500, 12480, 14.6, 1000, 3679, 34430, 12600, 14.6, 19.2, 
    4900, 826, 3679, 2100, 38000, 79, 11400, 21495, 3679, 200, 
    14.6, 100.6, 3679, 5300, 108.9, 3679, 2696, 7500, 171.6, 
    14.6, 99.2, 2452, 3679, 3218, 700, 69.7, 14.6, 91.5, 2452, 
    3679, 2900, 17572, 14.6, 14.6, 90.5, 2452, 49752, 3679, 1900, 
    14.6, 870, 85.2, 2452, 3679, 1600, 540, 14.6, 14.6, 79, 210, 
    2452, 28400, 720, 180, 420, 44289, 489, 3679, 840, 2900, 
    150, 870, 420, 14.6)), row.names = c(NA, -100L), class = "data.frame")

（我将from 和to 列转换为字符，因为它们有大量级别，否则输出会占用大量空间）

【问题讨论】：

为什么这个答案在这里不起作用？ stackoverflow.com/a/63689794/3962914 而不是 from 分组，你是为 to 做的？
是的，它可以工作，但之后我想以描述“来自”列的特征的方式将新列添加到数据中。如果我们只是按“to”分组并像在您的解决方案中那样进行计算，那么该新列将描述“to”列的特征。我基于“from”列创建新列，因此我的数据中的所有其他特征也描述了“from”列的特征。
虽然我们正在考虑to 列进行计算，但我认为这是您想要的，除非我感到困惑。你能检查一下输出吗，如果它不符合你的期望，你可以更新你的帖子，显示to列的输出和你的预期输出from？
我手动显示了输出将如何略高于数据的 dput 输出。你看到了吗？我根据“to”列计算了 total_trx_amount_received_in_last_6month 列，但根据“from”列添加了该信息。
这个怎么样？ data %>% mutate(amt = map2_dbl(from, date, ~sum(amount[to == .x & between(date, .y - 180, .y)]))) 。这是使用dplyr 和purrr。

标签： r date dplyr datatable sqldf

【解决方案1】：

我们可以使用map2_dbl 并选取amount 中的sum，它们位于6 个月的范围内。

library(dplyr)
library(purrr)

data %>% 
    mutate(amt = map2_dbl(from, date,
                ~sum(amount[to == .x & between(date, .y - 180, .y)])))

【讨论】：

它在小数据集上效果很好，但是当我在大数据集上尝试时，它似乎会永远运行
有没有有效的方法来做到这一点？
您可以将相同的代码转换为基本 R 和 data.table，但我不知道其他方法。
我尝试用future::plan(strategy = multisession)加速代码并将代码修改为data[, amt := furrr::future_pmap_dbl(list(from, date), ~sum(amount[to== .x & between(date, .y-180, .y)])) ]，但速度一点都没有提升。