如何每 15 分钟计算一次 "batch"

Question

1 2021-01-01 12:59:38
2 2021-01-01 14:08:59
3 2021-01-01 14:09:08
4 2021-01-01 14:11:30
5 2021-01-01 14:22:19
6 2021-01-01 14:41:07

我希望能够每 15 分钟计算一次条目数，但要在滚动的基础上进行。例如 12:59 将在 15 分钟内出现 1 个，14:08 => 14:22 将在 15 分钟内出现，因此这批 return 4 个，最后 14:41将在另一个 15 分钟的批次中单独出现。

我希望这是有道理的，并提前致谢。

抱歉没有包括这个

> dput(df)
structure(list(ClickedDate = structure(c(1609460198.707, 1609462979.593, 
1609465088.437, 1609476270.88, 1609478479.177, 1609479667.373, 
1609493081.887, 1609499187.29, 1609507506.37, 1609510989.533, 
1609511522.023, 1609511894.067, 1609512194.773, 1609512377.227, 
1609514474.153), tzone = "UTC", class = c("POSIXct", "POSIXt"
)), batch_no = c(1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 11L, 
12L, 12L, 12L, 13L), batch_size = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 3L, 3L, 3L, 1L)), row.names = c(NA, -15L), class = c("tbl_df", 
"tbl", "data.frame"))

新编辑 - 感谢您为此付出努力。我收到一个错误

Error in UseMethod("mutate") : 
  no applicable method for 'mutate' applied to an object of class "c('integer', 'numeric')"

这看起来很奇怪，我的变量在 class

> class(df$ClickedDate)
[1] "POSIXct" "POSIXt"

这是否适用于 mutate，或者我需要转换它吗？

> dput(df)
structure(list(ClickedDate = structure(c(1609460198.707, 1609462979.593, 
1609465088.437, 1609476270.88, 1609478479.177, 1609479667.373, 
1609493081.887, 1609499187.29, 1609507506.37, 1609510989.533, 
1609511522.023, 1609511894.067, 1609512194.773, 1609512377.227, 
1609514474.153), tzone = "UTC", class = c("POSIXct", "POSIXt"
)), batch_no = c(1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L, 11L, 
12L, 12L, 12L, 13L), batch_size = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 3L, 3L, 3L, 1L)), row.names = c(NA, -15L), class = c("tbl_df", 
"tbl", "data.frame"))

提前致谢

Answer 1

使用 runner 包将有助于这种情况。使用以下策略

library(tidyverse)
library(runner)

df %>% mutate(b_len = runner::runner(x = ClickedDate,
                             idx = ClickedDate,
                             k = "15 mins",
                             lag = "-14 mins",
                             f = length),
              b_no = purrr::accumulate(seq_len(length(b_len)-1), .init = b_len[1], ~ifelse(.x > .y, .x, .x + b_len[.x +1])),
              b_no = as.integer(as.factor(b_no))) %>%
  group_by(b_no) %>%
  mutate(b_len = n())

# A tibble: 15 x 3
# Groups:   b_no [12]
   ClickedDate         b_len  b_no
   <dttm>              <int> <int>
 1 2021-01-01 00:16:38     1     1
 2 2021-01-01 01:02:59     1     2
 3 2021-01-01 01:38:08     1     3
 4 2021-01-01 04:44:30     1     4
 5 2021-01-01 05:21:19     1     5
 6 2021-01-01 05:41:07     1     6
 7 2021-01-01 09:24:41     1     7
 8 2021-01-01 11:06:27     1     8
 9 2021-01-01 13:25:06     1     9
10 2021-01-01 14:23:09     2    10
11 2021-01-01 14:32:02     2    10
12 2021-01-01 14:38:14     3    11
13 2021-01-01 14:43:14     3    11
14 2021-01-01 14:46:17     3    11
15 2021-01-01 15:21:14     1    12

备注-

lag runner 函数中的参数允许向后时间 window （滚动）所以我使用负滞后来使用向前时间 window.
k runner 中的参数是给定的滚动长度 window
b_no 列最初标识 sliding/rolling window 最早 window 已用完，然后采用新的 window.
dense_rank 也可以使用（参见下面的替代方法）

或者

df %>% mutate(b_len = runner::runner(x = ClickedDate,
                             idx = ClickedDate,
                             k = "15 mins",
                             lag = "-14 mins",
                             f = length),
              b_no = purrr::accumulate(seq_len(length(b_len)-1), .init = b_len[1], ~ifelse(.x > .y, .x, .x + b_len[.x +1])),
              b_no = dense_rank(b_no)) %>%
  group_by(b_no) %>%
  mutate(b_len = n()) %>%
  ungroup()
# A tibble: 15 x 3
   ClickedDate         b_len  b_no
   <dttm>              <int> <int>
 1 2021-01-01 00:16:38     1     1
 2 2021-01-01 01:02:59     1     2
 3 2021-01-01 01:38:08     1     3
 4 2021-01-01 04:44:30     1     4
 5 2021-01-01 05:21:19     1     5
 6 2021-01-01 05:41:07     1     6
 7 2021-01-01 09:24:41     1     7
 8 2021-01-01 11:06:27     1     8
 9 2021-01-01 13:25:06     1     9
10 2021-01-01 14:23:09     2    10
11 2021-01-01 14:32:02     2    10
12 2021-01-01 14:38:14     3    11
13 2021-01-01 14:43:14     3    11
14 2021-01-01 14:46:17     3    11
15 2021-01-01 15:21:14     1    12

使用的数据

df
> df
# A tibble: 15 x 1
   ClickedDate        
   <dttm>             
 1 2021-01-01 00:16:38
 2 2021-01-01 01:02:59
 3 2021-01-01 01:38:08
 4 2021-01-01 04:44:30
 5 2021-01-01 05:21:19
 6 2021-01-01 05:41:07
 7 2021-01-01 09:24:41
 8 2021-01-01 11:06:27
 9 2021-01-01 13:25:06
10 2021-01-01 14:23:09
11 2021-01-01 14:32:02
12 2021-01-01 14:38:14
13 2021-01-01 14:43:14
14 2021-01-01 14:46:17
15 2021-01-01 15:21:14

如何每 15 分钟计算一次 "batch"

how to count a "batch" every 15 minutes

r

runner

batching

rolling-computation