在 tibble 列中找到最长的连续连胜

Question

我有以下问题：

my_tbl <- tribble(
  ~year, ~event_id, ~winner_id,
  2011,      "A",     4322,
  2011,      "A",     9604,
  2011,      "A",     1180,
  2013,      "A",     4322,
  2013,      "A",     9604,
  2013,      "A",     1663,
  2014,      "A",     4322,
  2016,      "A",     5478,
  2017,      "A",     4322,
  2017,      "A",     1663,
  2011,      "B",     4322,
  2013,      "B",     7893,
  2013,      "B",     1188,
  2014,      "B",     7893,
  2016,      "B",     2365,
  2017,      "B",     3407,
  2011,      "C",     5556,
  2013,      "C",     5556,
  2014,      "C",     1238,
  2016,      "C",     2391,
  2017,      "C",     2391,
  2011,      "D",     4219,
  2013,      "D",     7623,
  2014,      "D",     8003,
  2016,      "D",     2851,
  2017,      "D",     0418
)

我想按事件 ID 找出最多的连续获胜次数。我正在寻找的结果如下所示：

results_summary_tbl <- tribble(
  ~event_id, ~most_wins_in_a_row, ~number_of_winners, ~winners,                              ~years,
   "A",       3,                  1,                   "4322",                               "4322 = (2011, 2013, 2014)",
   "C",       2,                  2,                   "5556 , 2391",                        "5556 = (2011, 2013), 2391 = (2015, 2016)",
   "B",       2,                  1,                   "7893",                               "7893 = (2013, 2014)",
   "D",       1,                  5,                   "4219 , 7623 , 8003 , 2851 , 0418",   "4219 = (2011), 7623 = (2013), 8003 = (2014), 2851 = (2016), 0418 = (2017)"
)

请注意缺少年份，因为这些年份没有发生事件。

我提供了以下代码，但由于缺少年份而无法使用：

my_tbl %>% arrange(event_id, winner_id, year) %>%
  group_by(event_id, winner_id) %>%
  mutate(run = cumsum(year - lag(year, default = first(year)) > 1)) %>%
  count(event_id, winner_id, run) %>%
  group_by(event_id) %>%
  summarise(most_wins_in_a_row = max(n),
            number_of_winners = sum(n == most_wins_in_a_row),
            winners = paste0(winner_id[n == most_wins_in_a_row], collapse = ","))

Answer 1

我遇到了缺失年份的问题，所以我不得不使用带有 rle 的基本 R 方法来解决它。例如，我们取一个子集，其中 event_id=="A":

z = my_tbl[my_tbl$event_id =="A",]

要知道这个活动举办了多少年，谁赢了，我知道：

table(z$year,z$winner_id)

       1180 1663 4322 5478 9604
  2011    1    0    1    0    1
  2013    0    1    1    0    1
  2014    0    0    1    0    0
  2016    0    0    0    1    0
  2017    0    1    1    0    0

这简化了为每列查找最大连续 1 的问题。为此，我使用 rle,

apply(table(z$year,z$winner_id),2,function(i){
           k=rle(i)
           max(k$lengths[k$values == 1])
         })
1180 1663 4322 5478 9604 
   1    1    3    1    2

显示 4322 具有最长的连胜，因为 event_id == 一个子集。有了这个，很容易把你需要的输出写成一个data.frame。剩下的就是将这个函数应用于所有数据子集：

library(purrr)
library(tidyr)

findLongestStreak = function(z){
TAB=table(z$year,z$winner_id)
max_per_id = apply(TAB,2,function(i){
          k=rle(i)
          MAX = max(k$lengths[k$values == 1])
          YEARS = rownames(TAB)[which(rep(k$lengths==MAX,k$lengths) & i==1)[1:MAX]]
          data.frame(
            N=MAX,
            YEARS=paste(YEARS,collapse=","))
        })
max_per_id = do.call(rbind,max_per_id)
WINNERS = max_per_id$N == max(max_per_id$N)
data.frame(
  event_id = z$event_id[1],
  most_wins_in_a_row = max(max_per_id$N),
  number_of_winners = sum(WINNERS),
  winners=paste(rownames(max_per_id)[WINNERS],collapse=","),
  years=paste(max_per_id$YEARS[WINNERS],collapse=";"),
  stringsAsFactors=F
)
}

my_tbl %>% split(.$event_id) %>% map_dfr(findLongestStreak)
  event_id most_wins_in_a_row number_of_winners                 winners
1        A                  3                 1                    4322
2        B                  3                 1                    7893
3        C                  3                 2               2391,5556
4        D                  3                 5 418,2851,4219,7623,8003
                     years
1           2011,2013,2014
2                2013,2014
3      2016,2017;2011,2013
4 2017;2016;2011;2013;2014

在 tibble 列中找到最长的连续连胜

Finding the longest consecutive streak in a tibble column

r

tibble