返回所有可能的值组合

3

我有如下表:

Group Value
----  ----
1     A
1     B
1     C
1     D
2     A
2     B
2     C

对于这两个组,我想返回所有可能的值组合。例如对于第一组,可能的组合方式包括 (A,B), (A,C), (A,D), (B,C), (B,D), (C,D), (A,B,C), (B,D,C), (D,C,A), (C,A,B)。类似地,对于第二组,可能的组合方式为(A,B), (A,C), (B,C) [备注:我不想考虑仅有一个值、所有值和没有值的组合方式。因此,n个不同值将有2^(n)-n-1-1种组合方式]。
我想通过添加一个名为“Combi”的额外列来列出所有这些组合方式。该列按顺序编号不同的组合方式。
Group Combi Value
----  ----  ----
1     1     A
1     1     B
1     2     A
1     2     C
1     3     A
1     3     D
1     4     B
1     4     C
1     5     B
1     5     D
1     6     C
1     6     C
1     7     A
1     7     B
1     7     C
1     8     B
1     8     C
1     8     D
1     9     C
1     9     D
1     9     A
1     10    D
1     10    A
1     10    B
2     11    A
2     11    B
2     12    A
2     12    C
2     13    B
2     13    C

我该如何在R中完成这个任务?

4
您需要使用expand.grid函数。 - akrun
可能是[生成所有可能的排列(或n元组)]的重复问题(https://dev59.com/Xmox5IYBdhLWcg3wHAwA)。 - loki
3个回答

2
这里有一个通用的tidyverse解决方案,适用于具有多个项的值集。思路是使用combn(m = 2,3等)并将输出格式化为不同Groupm值的tibble。从那里,我们可以使用tidyverse函数map_dfrunnest。最后,由于我们有多个id而不是一个,我们构建一个唯一id的表格,构建唯一的combi id,并将其加入到结果中。
# convenience fonction to store combinations in a long format
combi_as_tibble <- function(n,values) combn(values,n) %>%
  {tibble(id = rep(seq(ncol(.)),each=nrow(.)),Value=c(.))}    
combi_as_tibble(2,letters[1:3]) # example
# # A tibble: 6 x 2
#      id Value
#   <chr> <chr>
# 1     1     a
# 2     1     b
# 3     2     a
# 4     2     c
# 5     3     b
# 6     3     c


df1 %>% group_by(Group) %>%
  summarize(combis = list(
    map_dfr(2:(length(unique(Value))-1),combi_as_tibble,Value,.id="id2")
  ))     %>% # by Group, build a long tibble with all combinations
  unnest %>% # unnest to get a long unnested table
  left_join(.,select(.,Group,id2,id) %>% distinct %>% mutate(combi=row_number())
  )      %>% # build combi ids
  select(Group,Value,combi) %>%
  as.data.frame

#    Group Value combi
# 1      1     A     1
# 2      1     B     1
# 3      1     A     2
# 4      1     C     2
# 5      1     A     3
# 6      1     D     3
# 7      1     B     4
# 8      1     C     4
# 9      1     B     5
# 10     1     D     5
# 11     1     C     6
# 12     1     D     6
# 13     1     A     7
# 14     1     B     7
# 15     1     C     7
# 16     1     A     8
# 17     1     B     8
# 18     1     D     8
# 19     1     A     9
# 20     1     C     9
# 21     1     D     9
# 22     1     B    10
# 23     1     C    10
# 24     1     D    10
# 25     2     A    11
# 26     2     B    11
# 27     2     A    12
# 28     2     C    12
# 29     2     B    13
# 30     2     C    13

数据

df1 <- read.table(text="Group Value
1     A
1     B
1     C
1     D
2     A
2     B
2     C",h=T,strin=F)

0
一个使用 data.table 的可能解决方案:
library(data.table)
setDT(dat)[, .(Value = {n <- 2:(uniqueN(Value)-1);
                        unlist(lapply(n, function(x) combn(Value, x)))})
           , by = Group
           ][, Combi := cumsum(c(1, diff(match(Value, LETTERS)) < 0))][]

这将会给出:

    Group Value Combi
 1:     1     A     1
 2:     1     B     1
 3:     1     A     2
 4:     1     C     2
 5:     1     A     3
 6:     1     D     3
 7:     1     B     4
 8:     1     C     4
 9:     1     B     5
10:     1     D     5
11:     1     C     6
12:     1     D     6
13:     1     A     7
14:     1     B     7
15:     1     C     7
16:     1     A     8
17:     1     B     8
18:     1     D     8
19:     1     A     9
20:     1     C     9
21:     1     D     9
22:     1     B    10
23:     1     C    10
24:     1     D    10
25:     2     A    11
26:     2     B    11
27:     2     A    12
28:     2     C    12
29:     2     B    13
30:     2     C    13

1
ABC不能成为第二组的排列,因为它包含了所有的项目。 - moodymudskipper

0

这是一个基本解决方案。内联注释。

#for each length (not incl. 1 and number of unique values), create all possible combinations using combn
combiLs <- by(dat, dat$Group, function(x) {
    #number of elements to choose excl. 1 and all values
    idx <- seq_along(x$Value)[-c(1, nrow(x))]

    do.call(rbind, lapply(idx, function(m) {
        #for each number of elements, generate all combinations
        sets <- combn(x$Value, m, simplify=FALSE)

        #get OP's desired format
        combi <- rep(seq_along(sets), each=m)
        data.frame(
            Group=x$Group[1],
            Combi=paste(x$Group[1], combi, sep="."),
            Value=unlist(sets))
    }))
})

#final output
do.call(rbind, combiLs)

另一种可能的 data.table 实现,同时使用 choosecombn

res <- setDT(dat)[, {
        idx <- seq_along(Value)[-c(1, .N)]
        list(
            Set=paste0(Group[1], ".", unlist(lapply(idx, function(m) rep(seq_len(choose(.N, m)), each=m)))),
            Value=unlist(lapply(idx, function(m) as.vector(combn(Value, m))))
        )
    }, by=Group]

res[, Combi := unlist(Map(rep, x=seq_along(rle(Set)$values), each=rle(Set)$lengths))]

数据:

dat <- data.frame(Group=c(rep(1,4), rep(2,3)), Value=c("A","B","C","D","A","B","C"))
dat

网页内容由stack overflow 提供, 点击上面的
可以查看英文原文,
原文链接