如何按季分组生成《老友记》角色同场相关性可视化图表？

核心问题原因 #

pairwise_cor（corrr包）默认对整个数据集计算相关性，group_by无法触发分组计算逻辑，因此会丢失season字段信息。需要手动按季拆分数据后分别计算，再整合结果。

具体实现步骤 #

1. 加载依赖包 #

library(tidyverse)
library(corrr)
library(ggplot2)

2. 数据预处理与按季拆分 #

假设你的数据集名为friends_script，包含season（季数）、episode（剧集）、character（角色）字段。先按季拆分数据集：

# 按季拆分，保留season字段
season_datasets <- friends_script %>%
  group_split(season, .keep = TRUE)

3. 批量计算各季角色同场相关性 #

遍历每个季的子数据集，先将数据转换为「每集-角色」的宽格式（标记角色是否在该集出现），再计算两两相关性并添加季数标记：

# 批量计算并整合结果
season_cor_results <- map_dfr(season_datasets, function(sub_df) {
  # 转换为宽格式：行=剧集，列=角色，值=是否出现（1=出现，0=未出现）
  episode_character_matrix <- sub_df %>%
    distinct(season, episode, character) %>%
    mutate(present = 1) %>%
    pivot_wider(
      names_from = character,
      values_from = present,
      values_fill = 0
    ) %>%
    select(-season, -episode) # 移除非角色字段
  
  # 计算两两相关性，添加当前季标记
  episode_character_matrix %>%
    pairwise_cor(method = "pearson") %>% # 可替换为"spearman"
    mutate(season = unique(sub_df$season))
})

4. 可视化方案 #

方案1：分面展示所有季的相关性

season_cor_results %>%
  filter(item1 != item2) %>% # 移除角色自身的相关性
  ggplot(aes(x = item1, y = item2, fill = correlation)) +
  geom_tile(color = "white") +
  scale_fill_gradient2(
    low = "#1a53ff",
    mid = "white",
    high = "#ff4d4d",
    midpoint = 0,
    limits = c(-1, 1)
  ) +
  labs(title = "老友记主卡司同场相关性（按季）", x = "", y = "") +
  theme_minimal() +
  theme(
    axis.text.x = element_text(angle = 45, hjust = 1),
    plot.title = element_text(hjust = 0.5)
  ) +
  facet_wrap(~season, ncol = 2) # 按季分面，可调整列数

方案2：单独生成并保存每一季的可视化图

# 遍历每个季，生成独立图表并保存
walk(unique(season_cor_results$season), function(season_num) {
  season_plot <- season_cor_results %>%
    filter(season == season_num, item1 != item2) %>%
    ggplot(aes(x = item1, y = item2, fill = correlation)) +
    geom_tile(color = "white") +
    scale_fill_gradient2(
      low = "#1a53ff",
      mid = "white",
      high = "#ff4d4d",
      midpoint = 0,
      limits = c(-1, 1)
    ) +
    labs(
      title = str_glue("老友记第{season_num}季主卡司同场相关性"),
      x = "", y = ""
    ) +
    theme_minimal() +
    theme(
      axis.text.x = element_text(angle = 45, hjust = 1),
      plot.title = element_text(hjust = 0.5)
    )
  
  # 保存图片到工作目录
  ggsave(
    filename = str_glue("friends_cor_season_{season_num}.png"),
    plot = season_plot,
    width = 8, height = 6, dpi = 300
  )
})

内容的提问来源于stack exchange，提问作者ECII