如何在R中的每个组中使变量元素彼此配对

Question

我有一个如下数据框：

data.frame(id = rep(1:2, each=4), 
           word = c('apple', 'pear', 'orange', 'banana',
                    'apple', 'watermellon','orange', 'grape') )

我需要将数据转换为以下表格。需要将每个组（id）中单词列中的所有单词配对，并创建两列word1和word2。

id    word1      word2
1     apple      pear
1     apple      orange
1     apple      banana
1     pear       orange
1     pear       banana
1     orange     banana
2     apple      watermellon
2     apple      Orange

Answer 1

我们可以按'id'进行分组，使用combn来获取'word'的成对组合，然后unnest进行输出

library(dplyr)
df1 %>%
  group_by(id) %>% 
  summarise(out = list(combn(word, 2, FUN = function(x)
     tibble(word1 = x[1], word2 = x[2]), simplify = FALSE))) %>%
  unnest %>%
  unnest
# A tibble: 12 x 3
#     id word1       word2      
#   <int> <fct>       <fct>      
# 1     1 apple       pear       
# 2     1 apple       orange     
# 3     1 apple       banana     
# 4     1 pear        orange     
# 5     1 pear        banana     
# 6     1 orange      banana     
# 7     2 apple       watermellon
# 8     2 apple       orange     
# 9     2 apple       grape      
#10     2 watermellon orange     
#11     2 watermellon grape      
#12     2 orange      grape     

---

或与data.table

library(data.table)
setDT(df1)[, as.data.frame(do.call(rbind, combn(as.character(word), 
     2, simplify = FALSE))), by = id] 

注意：这种用法combn仅在选择所需的组合而没有任何联接的情况下有效

数据

df1 <- data.frame(id = rep(1:2, each=4), 
           word = c('apple', 'pear', 'orange', 'banana',
                    'apple', 'watermellon','orange', 'grape') )

Answer 2

这是一个dplyr解决方案，它可以将数据框自身连接起来并删除不需要的对-

df %>% 
  inner_join(df, by = "id") %>% 
  filter(
    word.x != word.y &
    !duplicated(t(apply(., 1, sort)))
  ) %>% 
  rename(word1 = word.x, word2 = word.y)

   id       word1       word2
1   1       apple        pear
2   1       apple      orange
3   1       apple      banana
4   1        pear      orange
5   1        pear      banana
6   1      orange      banana
7   2       apple watermellon
8   2       apple      orange
9   2       apple       grape
10  2 watermellon      orange
11  2 watermellon       grape
12  2      orange       grape