在单独的列中每小时过去后获取字符列的模式？

Question

我有一个非常大的数据集 - 大约 2000 万个观测值，这是它的基本结构 -

           date       time      string
  1     01/01/2020   20:00:00     A  
  2     01/01/2020   20:13:12     B
  3     01/01/2020   20:37:45     C
  4     01/01/2020   20:39:07     D 
  5     01/01/2020   20:41:29     A
  6     01/01/2020   20:46:48     E
  7     01/01/2020   21:00:00     J

我想要一个新列，也许是“mode”，它会计算“string”列中最常出现的文本字符串，但仅以小时为间隔。所以 table 最终会变成这样 -

           date       time      string      mode
  1     01/01/2020   20:00:00     A          
  2     01/01/2020   20:13:12     B
  3     01/01/2020   20:37:45     C
  4     01/01/2020   20:39:07     D 
  5     01/01/2020   20:41:29     A
  6     01/01/2020   20:46:48     E
  7     01/01/2020   21:00:00     J          A
  8     01/01/2020   21:20:12     I
  9     01/01/2020   21:38:32     I  
  10    01/01/2020   21:43:12     A
  11    01/01/2020   21:49:50     I
  12    01/01/2020   21:54:50     B
  13    01/01/2020   22:00:00     A          I
  14    01/01/2020   22:03:45     B

因此每次小时数字递增时，都会对字符串列采用一种模式，然后模式测量会在下一个小时长的间隔内重置。

我是 R 的新手，所以很遗憾，我没有任何尝试或错误消息可以显示。我看过许多其他类似的主题/线程，但没有遇到任何帮助我找到可行解决方案的东西。当然不会要求任何人为我编写代码 - 任何建议都将不胜感激。

Answer 1

一个选项是使用 here

中的 Mode 函数

 Mode <- function(x) {
       ux <- unique(x)
     ux[which.max(tabulate(match(x, ux)))]
}

通过 paste 对 'date'、'time' 列创建分组变量，将其转换为具有 dmy_hms 的日期时间 class（来自 lubridate)，然后使用 ceiling_date 将“1 小时”指定为 unit，通过在 'string' 列上应用 Mode 创建 'mode' 列，然后使用case_when 到 return 该值仅在每个组的最后一行

library(dplyr)
library(lubridate
library(stringr)
df1 %>% 
     group_by(grp = ceiling_date(dmy_hms(str_c(date, time, sep=" ")),
          '1 hour')) %>% 
     mutate(mode = case_when(row_number() == n() ~ Mode(string), 
                TRUE  ~ "")) %>%
     ungroup %>% 
     select(-grp)

-输出

# A tibble: 14 x 4
#   date       time     string mode 
#   <chr>      <chr>    <chr>  <chr>
# 1 01/01/2020 20:04:01 A      ""   
# 2 01/01/2020 20:13:12 B      ""   
# 3 01/01/2020 20:37:45 C      ""   
# 4 01/01/2020 20:39:07 D      ""   
# 5 01/01/2020 20:41:29 A      ""   
# 6 01/01/2020 20:46:48 E      ""   
# 7 01/01/2020 21:00:00 J      "A"  
# 8 01/01/2020 21:20:12 I      ""   
# 9 01/01/2020 21:38:32 I      ""   
#10 01/01/2020 21:43:12 A      ""   
#11 01/01/2020 21:49:50 I      ""   
#12 01/01/2020 21:54:50 B      ""   
#13 01/01/2020 22:00:00 A      "I"  
#14 01/01/2020 22:03:45 B      "B"

数据

df1 <- structure(list(date = c("01/01/2020", "01/01/2020", "01/01/2020", 
"01/01/2020", "01/01/2020", "01/01/2020", "01/01/2020", "01/01/2020", 
"01/01/2020", "01/01/2020", "01/01/2020", "01/01/2020", "01/01/2020", 
"01/01/2020"), time = c("20:04:01", "20:13:12", "20:37:45", "20:39:07", 
"20:41:29", "20:46:48", "21:00:00", "21:20:12", "21:38:32", "21:43:12", 
"21:49:50", "21:54:50", "22:00:00", "22:03:45"), string = c("A", 
"B", "C", "D", "A", "E", "J", "I", "I", "A", "I", "B", "A", "B"
)), class = "data.frame", row.names = c("1", "2", "3", "4", "5", 
"6", "7", "8", "9", "10", "11", "12", "13", "14"))

Answer 2

使用library(data.table)我们可以做到

setDT(df1)[, hour := paste(date, sub(':.+','', time))]
df1[, n := seq(.N), by = .(hour, string)]
df1[, mode := string[which.max(n)], by=hour]

在单独的列中每小时过去后获取字符列的模式？

Getting the mode of a character column after every hour elapsed in separate column?

string

r

date

mode

数据