Grouping by ID, Grouping by time (within 5 minutes of each activity), Find Time Difference of Activity in R

Is there a way for R to group by ID, and then to identify a 'break' in time and then calculate time difference?
For instance:

                ID               TIME              
                A                12/18/2019 4:45:10 AM
                A                12/18/2019 4:45:11 AM
                A                12/18/2019 9:06:59 PM               
                B                12/18/2019 4:14:13 AM
                B                12/18/2019 4:14:14 AM

Does anyone know of a way to find the time duration for A? Notice this is not a difftime problem. I performed a certain activity at 4:45:10 am, then again at 4:45:11 am. I then stopped this activity, and picked back up at 9:06pm. Is there code that can accurately group IDs, and then group time whilst detecting a huge gap in the time to avoid inaccurate values?

This is not the correct solution.

                       diff<- data %>%
                       mutate(diff = difftime(as.POSIXct(Endtime, format = "%m/%d/%Y %I:%M:%S %p"), 
                       as.POSIXct(Starttime, format = "%m/%d/%Y %I:%M:%S %p"), units = "secs"))

Any help is greatly appreciated.
I will continue to research this. Thank you

Here's a way to do:

library(lubridate)
sample_df$TIME = mdy_hms(sample_df$TIME)

sample_df = sample_df %>%
            group_by(ID) %>%
            # lag basically bring the next value one step up
            # so we can subtract value at index 0 and index 1, index 1 and index 2 and so on....
            mutate(time_diff = TIME - lag(TIME, n = 1, default = NA)) %>% 
            mutate(time_diff = replace_na(time_diff, 0))

Hope this gives you some idea.
For understanding, do it in two steps:

sample_df = sample_df %>%
            group_by(ID) %>%
            mutate(time_lag = dplyr::lag(TIME, n = 1, default = NA)) %>% 
            mutate(time_diff = TIME - time_lag) %>% 
            mutate(time_diff = replace_na(time_diff, 0))

Check how time_lag column looks.

Like I mentioned above, the first thing to do is convert your date-times to a date-time object; I'm using lubridate for this. Since you want to keep delta within some threshold, I saved a threshold duration of 5 minutes which you can change as needed. If differences are more than that, make them NA.

I'm doing the diffing in 2 steps, just so you can see the original difference vs the one with long differences removed. You'll probably want to just do that in one step.

<!-- language-all: lang-r -->

library(dplyr)
library(lubridate)

thresh <- duration(5, units = "minutes")

sample_df %>%
  mutate(TIME = mdy_hms(TIME)) %>%
  group_by(ID) %>%
  mutate(diff1 = TIME - lag(TIME)) %>%
  mutate(delta = if_else(diff1 < thresh, diff1, NA_real_))
#> # A tibble: 10 x 4
#> # Groups:   ID [3]
#>    ID    TIME                diff1      delta  
#>    <chr> <dttm>              <drtn>     <drtn> 
#>  1 A     2019-12-18 04:45:10    NA secs NA secs
#>  2 A     2019-12-18 04:45:11     1 secs  1 secs
#>  3 A     2019-12-18 16:06:59 40908 secs NA secs
#>  4 A     2019-12-18 16:07:01     2 secs  2 secs
#>  5 B     2019-12-18 04:14:13    NA secs NA secs
#>  6 B     2019-12-18 04:14:14     1 secs  1 secs
#>  7 B     2019-12-18 04:14:15     1 secs  1 secs
#>  8 C     2019-12-18 04:59:49    NA secs NA secs
#>  9 C     2019-12-18 04:59:50     1 secs  1 secs
#> 10 C     2019-12-18 04:59:51     1 secs  1 secs

Using dplyr::if_else rather than the base ifelse was handy because it uses strict typing, which helped make sure I kept the delta column as a duration object, rather than losing its time component and just getting a numeric, which would be the case with NA instead of NA_real_.