R data.table Untergruppe einer Untermenge

Question

Ich habe einen Datensatz, der aus Kundentransaktionsdaten besteht. Es gibt Zeitstempel für bestimmte Ereignisse. Ich möchte nur Ereignisse erhalten, die vor einem bestimmten Ereignis stattgefunden haben, aber für jeden Kunden, nicht nur einen einzigen Zeitpunkt.

Hier ist ein Überblick:

 custId   date_time_recorded    event   dateTime1 
1 280512544 2012-11-13 15:25:37.947-08   shipped 2012-11-13 15:25:37 
2 280512544 2012-11-13 15:22:42.614-08  statusCheck 2012-11-13 15:22:42 
3 280512544 2012-11-13 15:03:16.62-08  statusCheck 2012-11-13 15:03:16 
4 280512544 2012-11-13 15:01:35.149-08  statusCheck 2012-11-13 15:01:35 
5 280512544 2012-11-13 14:45:41.964-08  status-picked 2012-11-13 14:45:41 
6 280512544 2012-11-13 14:44:57.664-08 warehouse_notified 2012-11-13 14:44:57 
7 280512544 2012-11-13 14:44:57.644-08  statusCheck 2012-11-13 14:44:57 
8 280512544 2012-11-13 13:05:15.725-08  recordCreated 2012-11-13 13:05:15 
9 280510610 2012-11-13 09:22:36.427-08   shipped 2012-11-13 09:22:36 
10 280510610 2012-11-13 09:20:07.202-08  statusCheck 2012-11-13 09:20:07 
11 280510610 2012-11-13 09:14:56.182-08  statusCheck 2012-11-13 09:14:56 

Ich mag würde nur Ereignisse erhalten, die vor dem ‚ausgeliefert‘ Ereignis passiert ist. Ich verwende derzeit ddply, um dies zu erreichen, aber es dauert eine lange Zeit.

keepPreShip <- function(x){ 
shipTime <- fastPOSIXct(x[grep("shipped", x$event, ignore.case = T), "date_time_recorded"],tz = "UTC") 
#shipTime <- fastPOSIXct(x[x$event =="shipped", "date_time_recorded"],tz = "UTC") 
x <- x[x$dateTime1 < shipTime,] 
} 

system.time(eventsMain1 <- ddply(ss1, .(custId), keepPreShip)) 

Gibt es einen schnelleren Weg, dies zu tun? Vielleicht mit data.table?

Hier ist ein dput der Daten:

> dput(ss1) 
structure(list(custId = c(280512544L, 280512544L, 280512544L, 
280512544L, 280512544L, 280512544L, 280512544L, 280512544L, 280510610L, 
280510610L, 280510610L, 280510610L, 280510610L, 280510610L, 280510610L, 
280510610L, 280511123L, 280511123L, 280511123L, 280511123L), 
    date_time_recorded = c("2012-11-13 15:25:37.947-08", "2012-11-13 15:22:42.614-08", 
    "2012-11-13 15:03:16.62-08", "2012-11-13 15:01:35.149-08", 
    "2012-11-13 14:45:41.964-08", "2012-11-13 14:44:57.664-08", 
    "2012-11-13 14:44:57.644-08", "2012-11-13 13:05:15.725-08", 
    "2012-11-13 09:22:36.427-08", "2012-11-13 09:20:07.202-08", 
    "2012-11-13 09:14:56.182-08", "2012-11-13 09:11:40.438-08", 
    "2012-11-13 09:03:51.571-08", "2012-11-13 09:03:51.461-08", 
    "2012-11-13 09:03:49.174-08", "2012-11-13 06:42:10.208-08", 
    "2012-11-13 13:51:05.039-08", "2012-11-13 13:13:16.452-08", 
    "2012-11-13 12:42:08.917-08", "2012-11-13 12:28:51.541-08" 
    ), event = c("shipped", "statusCheck", "statusCheck", "statusCheck", 
    "status-picked", "warehouse_notified", "statusCheck", "recordCreated", 
    "shipped", "statusCheck", "statusCheck", "statusCheck", "status-picked", 
    "warehouse_notified", "statusCheck", "recordCreated", "shipped", 
    "statusCheck", "statusCheck", "statusCheck"), dateTime1 = structure(c(1352820337.947, 
    1352820162.614, 1352818996.62, 1352818895.149, 1352817941.964, 
    1352817897.664, 1352817897.644, 1352811915.725, 1352798556.427, 
    1352798407.202, 1352798096.182, 1352797900.438, 1352797431.571, 
    1352797431.461, 1352797429.174, 1352788930.208, 1352814665.039, 
    1352812396.452, 1352810528.917, 1352809731.541), class = c("POSIXct", 
    "POSIXt"), tzone = "UTC")), .Names = c("custId", "date_time_recorded", 
"event", "dateTime1"), row.names = c(NA, 20L), class = "data.frame") 

Answer 1

Dies funktioniert. Ich hoffe es ist schneller.

require(data.table) 
# convert this column to date format first 
ss1$date_time_recorded <- as.POSIXct(ss1$date_time_recorded) 
dt <- data.table(ss1, key="custId") 
dt[, .SD[dateTime1 < date_time_recorded[event == "shipped"]], by=custId] 

#  custId date_time_recorded   event   dateTime1 
# 1: 280510610 2012-11-13 06:42:10 recordCreated 2012-11-13 06:42:10 
# 2: 280511123 2012-11-13 12:42:08 statusCheck 2012-11-13 12:42:08 
# 3: 280511123 2012-11-13 12:28:51 statusCheck 2012-11-13 12:28:51 
# 4: 280512544 2012-11-13 13:05:15 recordCreated 2012-11-13 13:05:15