在 R 中：根据事件的 ID 对事件序列进行分组，并计算第一个和最后一个看到的事件之间的时间差

Question

在下面的数据框中，我有可以重复的事件序列，直到生成新事件。 在选择这些相似的行后，我想使用一个函数来帮助我，计算 last seen event 的 timestamp_end 和 first eventtimestamp_start 之间的差异/strong>。

数据框：

DF1 <- data.frame(segment_id = c(1, 1, 1, 1, 2 , 3, 4), first_event= c("a", "a", "a","a", "a", "b","c" ), second_event = c("a", "a","a","a", "b", "c", "c"), timestamp_start = c("2019-06-06 11:47:00","2019-06-06 12:59:38", "2019-06-06 13:01:03", "2019-06-06 14:47:03   ", "2019-06-06 18:47:00", "2019-06-06 22:47:00", "2019-06-07 02:47:00") , timestamp_end = c("2019-06-06 12:59:38", "2019-06-06 13:01:03", "2019-06-06 14:47:03", "2019-06-06 18:47:00", "2019-06-06 22:47:00    ", "2019-06-07 02:47:00", "2019-06-07 06:47:00"))


segment_id   first_event   second_event          timestamp_start            timestamp_end

  1            a              a                2019-06-06 11:47:00        2019-06-06 12:59:38
  1            a              a                2019-06-06 12:59:38        2019-06-06 13:01:03
  1            a              a                2019-06-06 13:01:03        2019-06-06 14:47:03
  1            a              a                2019-06-06 14:47:03        2019-06-06 18:47:00
  2            a              b                2019-06-06 18:47:00        2019-06-06 22:47:00
  3            b              c                2019-06-06 22:47:00        2019-06-07 02:47:00
  4            c              c                2019-06-07 02:47:00        2019-06-07 06:47:00

所以，我尝试了 dplyrpackage 和 group_by() 和 mutate() 函数。但是，我不确定哪个函数可以帮助我获得持续时间。

DF2 <- DF1 %>%
   group_by(segment_id)%>%
   mutate("duration" = difftime(????) , units = 'hours')

我正在寻找 DF2 的最终结果应该是这样的：

>DF2
segment_id   first_event   second_event          timestamp_start            timestamp_end        duration

  1            a              a                2019-06-06 11:47:00        2019-06-06 18:47:00      7        
  2            a              b                2019-06-06 18:47:00        2019-06-06 22:47:00      4
  3            b              c                2019-06-06 22:47:00        2019-06-07 02:47:00      4
  4            c              c                2019-06-07 02:47:00        2019-06-07 06:47:00      4

感谢您在这方面的帮助。

Answer 1

使用data.table的选项：

setDT(DF1)[, .(timestamp_start=min(timestamp_start), timestamp_end=max(timestamp_end)), .(segment_id, first_event, second_event)][, 
    duration := difftime(timestamp_end, timestamp_start, units="hours")][]

数据：

library(data.table)
cols <- c("timestamp_start", "timestamp_end")
setDT(DF1)[, (cols) := lapply(.SD, as.POSIXct, format="%Y-%m-%d %T"), .SDcols=cols]

Answer 2

library(lubridate)

DF1 %>%
    mutate_at(vars(timestamp_start, timestamp_end)
              , function(x) ymd_hms(as.character(x))) %>%
    group_by(segment_id) %>%
    summarise(first_event = first(first_event)
              , second_event = last(second_event)
              , timestamp_start = first(timestamp_start)
              , timestamp_end = last(timestamp_end)
              , duration = max(timestamp_end) - min(timestamp_start))

  segment_id first_event second_event timestamp_start     timestamp_end       duration
       <dbl> <fct>       <fct>        <dttm>              <dttm>              <drtn>  
1          1 a           a            2019-06-06 11:47:00 2019-06-06 18:47:00 7 hours 
2          2 a           b            2019-06-06 18:47:00 2019-06-06 22:47:00 4 hours 
3          3 b           c            2019-06-06 22:47:00 2019-06-07 02:47:00 4 hours 
4          4 c           c            2019-06-07 02:47:00 2019-06-07 06:47:00 4 hours 

Answer 3

我们可以将timestamp_start和timestamp_end列更改为POSIXct类型，group_bysegment_id，first_event和second_event，得到firsttimestamp_start和@98765343@30@98765343@30并计算它们之间的小时差。

library(dplyr)

DF1 %>%
  mutate_at(vars(starts_with('timestamp')), as.POSIXct) %>%
  group_by(segment_id, first_event, second_event) %>%
  summarise(timestamp_start = first(timestamp_start), 
            timestamp_end = last(timestamp_end),
            duration = as.numeric(difftime(timestamp_end, 
                      timestamp_start, units = "hours")))

#     segment_id first_event second_event timestamp_start     timestamp_end       duration
#       <dbl> <fct>       <fct>        <dttm>              <dttm>                 <dbl>
#1          1 a           a            2019-06-06 11:47:00 2019-06-06 18:47:00        7
#2          2 a           b            2019-06-06 18:47:00 2019-06-06 22:47:00        4
#3          3 b           c            2019-06-06 22:47:00 2019-06-07 02:47:00        4
#4          4 c           c            2019-06-07 02:47:00 2019-06-07 06:47:00        4