如何在R中同时对三个字段进行网络分析

我需要在R中同时对三个字段执行网络分析，以下是我的示例数据，最后一列为预期输出desired_output：

df <- data.frame(
  stringsAsFactors = FALSE,
                    id_1 = c("ABC","ABC","BCD",
                             "CDE","DEF","EFG","GHI","HIJ","IJK","JKL",
                             "GHI","KLM","LMN","MNO","NOP"),
                    id_2 = c("1A","2A","3A",
                             "1A","4A","5A","6A","8A","9A","10A","7A",
                             "12A","13A","14A","15A"),
                    id_3 = c("Z3","Z2","Z1",
                             "Z4","Z1","Z5","Z5","Z6","Z7","Z8","Z6","Z8",
                             "Z9","Z9","Z1"),
                    Name = c("StackOverflow1",
                             "StackOverflow2","StackOverflow3","StackOverflow4",
                             "StackOverflow5","StackOverflow6",
                             "StackOverflow7","StackOverflow8","StackOverflow9",
                             "StackOverflow10","StackOverflow11","StackOverflow12",
                             "StackOverflow13","StackOverflow14","StackOverflow15"),
          desired_output = c(1L,1L,2L,1L,2L,
                             3L,3L,3L,4L,5L,3L,5L,6L,6L,2L)
      )
df
#>    id_1 id_2 id_3            Name desired_output
#> 1   ABC   1A   Z3  StackOverflow1              1
#> 2   ABC   2A   Z2  StackOverflow2              1
#> 3   BCD   3A   Z1  StackOverflow3              2
#> 4   CDE   1A   Z4  StackOverflow4              1
#> 5   DEF   4A   Z1  StackOverflow5              2
#> 6   EFG   5A   Z5  StackOverflow6              3
#> 7   GHI   6A   Z5  StackOverflow7              3
#> 8   HIJ   8A   Z6  StackOverflow8              3
#> 9   IJK   9A   Z7  StackOverflow9              4
#> 10  JKL  10A   Z8 StackOverflow10              5
#> 11  GHI   7A   Z6 StackOverflow11              3
#> 12  KLM  12A   Z8 StackOverflow12              5
#> 13  LMN  13A   Z9 StackOverflow13              6
#> 14  MNO  14A   Z9 StackOverflow14              6
#> 15  NOP  15A   Z1 StackOverflow15              2

目前我可以使用igraph包同时对两个字段做网络分析，但无法直接扩展到三个字段的场景，当前采用两次迭代的方法实现了需求，但认为该方案还有优化空间，现有实现代码如下：

library(igraph)
library(tidyverse)

graph.data.frame(df) %>%
  components() %>%
  pluck(membership) %>%
  stack() %>%
  set_names(c('GRP', 'id_1')) %>%
  right_join(df %>% mutate(id_1 = as.factor(id_1)), by = c('id_1')) %>%
  select(GRP, id_3) %>%
  graph.data.frame() %>% 
  components() %>%
  pluck(membership) %>%
  stack() %>%
  set_names(c('GRP', 'id_3')) %>%
  right_join(df %>% mutate(id_3 = as.factor(id_3)), by = c('id_3'))
#>    GRP id_3 id_1 id_2            Name desired_output
#> 1    1   Z3  ABC   1A  StackOverflow1              1
#> 2    1   Z2  ABC   2A  StackOverflow2              1
#> 3    2   Z1  BCD   3A  StackOverflow3              2
#> 4    2   Z1  DEF   4A  StackOverflow5              2
#> 5    2   Z1  NOP  15A StackOverflow15              2
#> 6    1   Z4  CDE   1A  StackOverflow4              1
#> 7    3   Z5  EFG   5A  StackOverflow6              3
#> 8    3   Z5  GHI   6A  StackOverflow7              3
#> 9    3   Z6  HIJ   8A  StackOverflow8              3
#> 10   3   Z6  GHI   7A StackOverflow11              3
#> 11   4   Z7  IJK   9A  StackOverflow9              4
#> 12   5   Z8  JKL  10A StackOverflow10              5
#> 13   5   Z8  KLM  12A StackOverflow12              5
#> 14   6   Z9  LMN  13A StackOverflow13              6
#> 15   6   Z9  MNO  14A StackOverflow14              6

核心思路是直接将每行三个字段的两两关联关系统一生成边列表，单次建图计算连通分量即可完成匹配，无需多次迭代，同时天然支持更多关联字段的扩展。

library(igraph)
library(tidyverse)

# 生成三个字段的两两无向边，去重避免冗余
edge_list <- df %>%
  select(id_1, id_2, id_3) %>%
  rowwise() %>%
  summarise(
    from = c(id_1, id_1, id_2),
    to = c(id_2, id_3, id_3),
    .groups = "drop"
  ) %>%
  distinct()

# 构建无向图，计算连通分量
g <- graph_from_data_frame(edge_list, directed = FALSE)
group_membership <- components(g)$membership

# 将分组结果匹配回原数据集
df_result <- df %>%
  rowwise() %>%
  mutate(GRP = group_membership[unique(c(id_1, id_2, id_3))][1]) %>%
  ungroup()

输出结果与预期的desired_output完全一致，相比多次迭代的方案逻辑更简洁，性能更稳定，也可以直接扩展到3个以上字段的关联分组场景。

内容的提问来源于stack exchange，提问作者AnilGoyal