基于R ggplot2实现矩阵上下三角分示accuracy与speed的热力图

核心思路 #

通过数据筛选拆分上下三角区域，分别映射speed和accuracy变量，再添加单元格对角线实现视觉分隔，同时保证两个指标的热力色标适配各自数值范围。

1. 依赖包加载 #

library(ggplot2)
# 若需额外数据格式转换可加载reshape2，本例直接用原宽表即可
# library(reshape2)

2. 数据预处理 #

假设你的数据集名为df，包含col1、col2、accuracy、speed字段。先给每行标记所属区域，并提取对应展示数值：

# 标记区域：上三角/下三角/对角线
df$region <- ifelse(df$col1 < df$col2, "upper", 
                    ifelse(df$col1 > df$col2, "lower", "diag"))

# 提取对应区域的展示值：上三角用speed，下三角用accuracy，对角线设为空
df$value <- ifelse(df$region == "upper", df$speed,
                   ifelse(df$region == "lower", df$accuracy, NA_real_))

若已安装dplyr，用case_when替代嵌套ifelse会更清晰：

library(dplyr)
df <- df %>%
  mutate(region = case_when(
    col1 < col2 ~ "upper",
    col1 > col2 ~ "lower",
    TRUE ~ "diag"
  ),
  value = case_when(
    region == "upper" ~ speed,
    region == "lower" ~ accuracy,
    TRUE ~ NA_real_
  ))

3. 最终绘图代码 #

# 初始化绘图对象，分图层绘制上下三角+对角线
p <- ggplot() +
  # 下三角：accuracy热力图
  geom_tile(data = df[df$region == "lower", ],
            aes(x = col2, y = col1, fill = accuracy),
            color = "white") +
  # 上三角：speed热力图
  geom_tile(data = df[df$region == "upper", ],
            aes(x = col2, y = col1, fill = speed),
            color = "white") +
  # 绘制每个单元格的左下→右上对角线（视觉分隔）
  geom_segment(data = df,
               aes(x = as.numeric(col2) - 0.5, y = as.numeric(col1) - 0.5,
                   xend = as.numeric(col2) + 0.5, yend = as.numeric(col1) + 0.5),
               color = "black", size = 0.5) +
  # 对角线区域设为空白（可自定义填充色）
  geom_tile(data = df[df$region == "diag", ],
            aes(x = col2, y = col1),
            fill = "white", color = "white") +
  # 自定义双指标色标：区分accuracy（蓝系）和speed（红系）
  scale_fill_gradientn(colors = c("#2A9D8F", "#FFFFFF", "#E76F51"),
                       values = c(0, 0.5, 1),
                       breaks = c(min(df$accuracy, na.rm = T), max(df$accuracy, na.rm = T),
                                  min(df$speed, na.rm = T), max(df$speed, na.rm = T)),
                       labels = c(paste("Accuracy:", round(min(df$accuracy, na.rm = T), 2)),
                                  paste("Accuracy:", round(max(df$accuracy, na.rm = T), 2)),
                                  paste("Speed:", round(min(df$speed, na.rm = T), 0)),
                                  paste("Speed:", round(max(df$speed, na.rm = T), 0)))) +
  # 反转y轴，让下三角显示在左下角（符合常规矩阵阅读习惯）
  scale_y_discrete(limits = rev(levels(df$col1))) +
  # 标题与标签设置
  labs(title = "Accuracy (Lower) & Speed (Upper) Heatmap",
       x = "Column 2", y = "Column 1", fill = "Metric Value") +
  # 主题优化
  theme_minimal() +
  theme(panel.grid = element_blank(),
        axis.text.x = element_text(angle = 45, hjust = 1))

# 输出图像
print(p)

关键细节说明 #

• 分图层绘制上下三角：避免speed和accuracy数值范围差异导致的色标失真
• 对角线绘制：通过将离散的col1/col2转为数值，偏移0.5定位到热力块的边界，实现精准对角线
• 色标自定义：可根据需求调整颜色方案，若需完全独立的双图例，可使用ggnewscale包实现

内容的提问来源于stack exchange，提问作者Dovini Jayasinghe