根据条件在 R 中使用 data.table 合并两行的值

Question

在我有一个 activity、开始和停止时间以及一个 ID 的数据集中，我想合并同一列中两行的值，并在多个条件适用时更新其他列。首先是一个数据示例：

library(data.table)
DT <- data.table(person=c(1,1,1,1,2,2,2,2,2,3,3,3,3),
             activity=c("grab", "walk", "remove", "delete", "run", "talk", "walk", "remove",
                        "grab", "walk", "delete", "talk", "remove"),
             start_time=c(0,1,3,6,0,2,2,3,3,3,6,6,7), stop_time=c(1,3,5,7,1,4,4,8,4,5,7,7,8))
 DT

我想为每个人更新开始和停止时间并合并列 'activity'，如果：

活动并行进行。具体来说，如果后面activity的start_time在前面activityactivity的stop_time之前。或者：
如果一个人的活动开始时间或结束时间相同。

更新的行应反映组合 activity 的开始和停止时间，除更新的行外，所有行都应删除。以下是我希望通过我提供的数据样本实现的目标：

DT.goal <- data.table(person=c(1,1,2,2,3,3),
                  activity=c("grab + walk + remove", "delete", "run", "talk + walk + grab + remove",
                             "walk", "delete + talk + remove"),
                  start_time=c(0,6,0,2,3,6), stop_time=c(5,7,1,8,5,8))
DT.goal

到目前为止，我已经进行了以下不完整的尝试：

DT.test <- DT[start_time <= shift(stop_time, 1L, type="lag"), 
          cond := T, by=person]
DT.test <- DT.test[cond==T, 
          new_activity := paste(activity, shift(activity, 1L, type="lag")), by=person]
DT.test <- DT.test[, new_start := start_time, by=person][cond==T, new_start := min(start_time), by=person]
DT.test <- DT.test[, new_stop := stop_time, by=person][cond==T, new_stop := max(stop_time), by=person]

但是，使用 shift(, type="lag) 对于每个人的第一行不是很有用，因为它没有前一行可供查看。此外，如果条件未计算为 TRUE，paste() 将粘贴 NA。

有人可以帮我吗？

Answer 1

请检查以下内容：

library(data.table)
DT <- data.table(person=c(1,1,1,1,2,2,2,2,2,3,3,3,3),
                 activity=c("grab", "walk", "remove", "delete", "run", "talk", "walk", "remove",
                            "grab", "walk", "delete", "talk", "remove"),
                 start_time=c(0,1,3,6,0,2,2,3,3,3,6,6,7), stop_time=c(1,3,5,7,1,4,4,8,4,5,7,7,8))

setorder(DT, person, start_time)
DT[, concatenate := start_time %in% stop_time | stop_time %in% start_time | duplicated(start_time) | duplicated(start_time, fromLast=TRUE), by = "person"]
DT[, concatenate_grp := rleid(concatenate), by = "person"]
DT[, paste(activity, collapse = " + "), by = c("person", "concatenate_grp")]
DT.goal <- DT[, .(activity = paste(activity, collapse = " + "), start_time = min(start_time), stop_time = max(stop_time)), by = c("person", "concatenate_grp")][, concatenate_grp := NULL]

这导致：

   person                    activity start_time stop_time
1:      1        grab + walk + remove          0         5
2:      1                      delete          6         7
3:      2                         run          0         1
4:      2 talk + walk + remove + grab          2         8
5:      3                        walk          3         5
6:      3      delete + talk + remove          6         8

Answer 2

使用dplyr，我们arrange数据person和开始和停止时间。我们创建一个 group，其中每组中每个 person 和 select 的时间重叠 first start_time 和 last stop_time并将每个组中的所有活动串联起来。

library(dplyr)

DT %>%
  arrange(person, start_time, stop_time) %>%
  group_by(person, group = cumsum(start_time > 
                lag(stop_time, default = first(stop_time)))) %>%
  summarise(start_time = first(start_time), 
            stop_time = last(stop_time), 
            activity = paste(activity, collapse = " + ")) %>%
  select(-group)

#  person start_time stop_time activity                   
#   <dbl>      <dbl>     <dbl> <chr>                      
#1      1          0         5 grab + walk + remove       
#2      1          6         7 delete                     
#3      2          0         1 run                        
#4      2          2         8 talk + walk + grab + remove
#5      3          3         5 walk                       
#6      3          6         8 delete + talk + remove

Answer 3

setorder(DT, person, stop_time)
DT[, 
   break_here := start_time > shift(stop_time, 1, stop_time[1]) , 
   by = person
   ][, 
     .(activity = paste(activity, collapse = " + "), start_time = start_time[1], stop_time = stop_time[.N]), 
     keyby = .(person, helper_var = cumsum(break_here))
     ][, !"helper_var"]


   person                    activity start_time stop_time
1:      1        grab + walk + remove          0         5
2:      1                      delete          6         7
3:      2                         run          0         1
4:      2 talk + walk + grab + remove          2         8
5:      3                        walk          3         5
6:      3      delete + talk + remove          6         8

Answer 4

另一种选择是借鉴 David Aurenburg 的解决方案

setorder(DT, person, start_time, stop_time)
DT[, g := c(0L, cumsum(shift(start_time, -1L) > cummax(stop_time))[-.N]), person]
DT[, .(activity=paste(activity, collapse=" + "), 
        start_time=min(start_time), stop_time=max(stop_time)), 
    .(person, g)]

输出：

   person g                    activity start_time stop_time
1:      1 0        grab + walk + remove          0         5
2:      1 1                      delete          6         7
3:      2 0                         run          0         1
4:      2 1 talk + walk + grab + remove          2         8
5:      3 0                        walk          3         5
6:      3 1      delete + talk + remove          6         8

根据条件在 R 中使用 data.table 合并两行的值

Merging values of two rows based on conditions using data.table in R

r

event-log

data.table