Data.table:加入ID和Date键,但希望在第一个表中的日期键之前(或等于)最接近的日期

Question

我真的很感激这个问题的一些帮助,我无法在SO上找到足够接近的例子.

我有两个data.tables,第一个叫做customer.table,包含特定时间戳(AsOfDate)的成员快照,第二个表activity.table用于描述发送给该客户的营销活动ActivityDate.

我想找到客户数据表中每条记录的AsOfDate之前或之前发送给成员的最新ActivityDate(即最长日期).

我已经看了几个问题(一个接近的问题是:处理一个ID重复的表),但我不确定如何将条件(ActivityDate <AsOfDate)与活动日期的最大值组合 - 我还希望保留连接中两个表的所有列,因为我需要计算ActivityDate和AsOfDate之间的时间.我仍然没有时间使用roll ...

#libraries
library(lubridate)
library(data.table)

#data
customer.table = structure(list(CustomerID = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 
4), AsOfDate = structure(c(1435622400, 1435622400, 1435622400, 
1435622400, 1435622400, 1435622400, 1435622400, 1435622400, 1435622400, 
1435622400, 1394150400), tzone = "UTC", class = c("POSIXct", 
"POSIXt")), distance = c(2.17380476584343, 29.4024827688224, 
3.01353310956009, 18.4923143452557, 294.878606580665, 11.8870209430565, 
9.54438580030996, 24.2192034858273, 15.0069335290262, 10.4513664447137, 
18.4923143452557)), .Names = c("CustomerID", "AsOfDate", "distance"
), row.names = c("1", "5", "8", "10", "18", "28", "33", "37", 
"45", "47", "101"), class = "data.frame")

activity.table = structure(list(CustomerID = c(3, 5, 8, 10, 4, 10, 2, 2, 5, 7,
5, 8, 4, 6, 10, 6, 5, 4, 2, 5, 5, 6, 5, 5, 10, 8, 6, 4, 5, 8,
7, 1, 8, 10, 7, 8, 4, 1, 1, 10, 9, 7, 4, 6, 9, 10, 8, 3, 5, 8,
1, 4, 4), ActivityDate = structure(c(1330560000, 1368144000,
1332855900, 1337817600, 1370822400, 1365984000, 1337817600, 1368144000,
1331164800, 1331164800, 1394150400, 1394150400, 1396224000, 1393891200,
1393891200, 1398643200, 1396310400, 1399334400, 1399939200, 1403222400,
1402358400, 1404086400, 1425254400, 1426464000, 1426464000, 1426464000,
1427155200, 1429056000, 1429056000, 1429056000, 1363737600, 1332201600,
1330560000, 1433116800, 1433289600, 1433289600, 1338462000, 1366628400,
1335885300, 1427241600, 1427241600, 1427241600, 1430265600, 1430265600,
1430265600, 1430265600, 1365503400, 1338394200, 1430265600, 1430265600,
1432598400, 1433894400, 1426723200), tzone = "UTC", class = c("POSIXct",
"POSIXt")), row.index = 1:53), .Names = c("CustomerID", "ActivityDate",
"row.index"), row.names = c(NA, -53L), class = "data.frame")

 # what does the data look like
> head(activity.table)
  CustomerID        ActivityDate row.index
1          3 2012-03-01 00:00:00         1
2          5 2013-05-10 00:00:00         2
3          8 2012-03-27 13:45:00         3
4         10 2012-05-24 00:00:00         4
5          4 2013-06-10 00:00:00         5
6         10 2013-04-15 00:00:00         6
> head(customer.table)
   CustomerID   AsOfDate   distance
1           1 2015-06-30   2.173805
5           2 2015-06-30  29.402483
8           3 2015-06-30   3.013533
10          4 2015-06-30  18.492314
18          5 2015-06-30 294.878607
28          6 2015-06-30  11.887021

谢谢你的协助.

Answer 1

看来你正在寻找一个简单的滚动连接.首先,我们将转换为data.table对象(请注意我在CRAN上使用最新版本的此解决方案(V 1.9.6+)

library(data.table) # V 1.9.6+
setDT(customer.table)
setDT(activity.table)

然后,对于每一行,customer.table我们将尝试activity.table在滚动到无穷大时加入最接近的值

indx <- activity.table[customer.table, 
                       on = c(CustomerID = "CustomerID",
                              ActivityDate = "AsOfDate"), 
                       roll = Inf,
                       which = TRUE]

indx
# [1] 51 19 48 52 49 44 35 36 45 34  5

indx是activity.table最接近每行的日期的位置向量customer.table.

现在,剩下的就是加入回来了 customer.table

customer.table[, MostRecentDate := activity.table[indx,ActivityDate]]
customer.table
#     CustomerID   AsOfDate   distance      MostRecentDate
#  1:          1 2015-06-30   2.173805 2015-05-26 00:00:00
#  2:          2 2015-06-30  29.402483 2014-05-13 00:00:00
#  3:          3 2015-06-30   3.013533 2012-05-30 16:10:00
#  4:          4 2015-06-30  18.492314 2015-06-10 00:00:00
#  5:          5 2015-06-30 294.878607 2015-04-29 00:00:00
#  6:          6 2015-06-30  11.887021 2015-04-29 00:00:00
#  7:          7 2015-06-30   9.544386 2015-06-03 00:00:00
#  8:          8 2015-06-30  24.219203 2015-06-03 00:00:00
#  9:          9 2015-06-30  15.006934 2015-04-29 00:00:00
# 10:         10 2015-06-30  10.451366 2015-06-01 00:00:00
# 11:          4 2014-03-07  18.492314 2013-06-10 00:00:00