R中含大量类别变量的热力图优化绘制技术问询

问题背景 #

刚接触R语言，尝试用热力图对比职业（30+类别）与受伤类型（30+类别）两个分类变量时，因类别过多导致热力图及坐标轴杂乱，无法有效解读结果。

示例数据与原代码 #

示例数据：

# 职业变量（实际含30+类别）
OCCUPATION <- c("Warehouse Bagger","Maintanence Man","Utility Man","Errand Boy") # 另有20+类别
# 受伤类型变量（实际含30+类别）
NATURE_INJURY <- c("Cut","Sprain","Bruise","Burn","Damaged Lung") # 另有30+类别

原绘制代码：

library(dplyr)
counting <- count(new_data4, OCCUPATION, NATURE_INJURY)
ggplot(data = counting, mapping = aes(x = OCCUPATION,
                               y = NATURE_INJURY)) +
    geom_tile(mapping = aes(fill = n))

优化方案 #

1. 合并低频类别 #

将计数占比极低的职业或受伤类型合并为「其他」类别，聚焦核心群体，直接减少类别数量：

# 保留计数前20的职业，其余归为"其他"
top_occupations <- new_data4 %>% 
  count(OCCUPATION) %>% 
  arrange(desc(n)) %>% 
  slice(1:20) %>% 
  pull(OCCUPATION)

# 保留计数前20的受伤类型，其余归为"其他"
top_injuries <- new_data4 %>% 
  count(NATURE_INJURY) %>% 
  arrange(desc(n)) %>% 
  slice(1:20) %>% 
  pull(NATURE_INJURY)

# 生成清洗后的数据
clean_data <- new_data4 %>% 
  mutate(
    OCCUPATION = ifelse(OCCUPATION %in% top_occupations, OCCUPATION, "其他"),
    NATURE_INJURY = ifelse(NATURE_INJURY %in% top_injuries, NATURE_INJURY, "其他")
  )

# 重新统计并绘图
counting_clean <- count(clean_data, OCCUPATION, NATURE_INJURY)
ggplot(counting_clean, aes(x = OCCUPATION, y = NATURE_INJURY)) +
  geom_tile(aes(fill = n))

2. 优化坐标轴布局 #

通过旋转标签、交换坐标轴避免重叠，提升静态图可读性：

# 旋转x轴标签，避免横向重叠
ggplot(counting, aes(x = OCCUPATION, y = NATURE_INJURY)) +
  geom_tile(aes(fill = n)) +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))

# 交换坐标轴（适合两类别数量差异较大的场景）
ggplot(counting, aes(x = OCCUPATION, y = NATURE_INJURY)) +
  geom_tile(aes(fill = n)) +
  coord_flip()

3. 过滤低计数组合 #

只保留计数大于阈值的职业-受伤类型组合，去掉无意义的空白/极低计数单元格：

# 过滤计数≥5的组合（阈值可根据数据调整）
counting_filtered <- counting %>% filter(n >= 5)

ggplot(counting_filtered, aes(x = OCCUPATION, y = NATURE_INJURY)) +
  geom_tile(aes(fill = n)) +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))

4. 改用交互式热力图 #

借助plotly包生成交互式图，鼠标悬停可查看具体类别与计数，无需压缩坐标轴：

library(plotly)

p <- ggplot(counting, aes(x = OCCUPATION, y = NATURE_INJURY, fill = n)) +
  geom_tile() +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1))

# 转换为交互式热力图
ggplotly(p)

内容的提问来源于stack exchange，提问作者HU EW