在RStudio中使用geom_tile生成空白热力图的问题排查

问题描述 #

我是R编程新手，尝试可视化不同聚类算法得到的簇之间的Normalised Mutual Information（归一化互信息）。目标是生成类似ggcorrplot样式的热力图，但由于数据并非相关系数，ggcorrplot无法使用，因此尝试用ggplot2手动绘制。

手动创建的数据框如下：

Nmi_frame <- data.frame(
  "mesure1" = c("fg","fg","fg","fg","fg","fg","fg",
                "sp","sp","sp","sp","sp","sp","sp",
                "wt","wt","wt","wt","wt","wt","wt",
                "lv","lv","lv","lv","lv","lv","lv",
                "sg","sg","sg","sg","sg","sg","sg",
                "ca","ca","ca","ca","ca","ca","ca",
                "cf","cf","cf","cf","cf","cf","cf"),

  "mesure2"= c( "fg","sp","wt","lv","sg","ca","cf",
                 "fg","sp","wt","lv","sg","ca","cf",
                 "fg","sp","wt","lv","sg","ca","cf",
                 "fg","sp","wt","lv","sg","ca","cf",
                 "fg","sp","wt","lv","sg","ca","cf",
                 "fg","sp","wt","lv","sg","ca","cf",
                 "fg","sp","wt","lv","sg","ca","cf"),

  "nmi_value" = c(1,0.3671792,0.3260069,0.2499102,0.2410074,0.2586295,0.1722615,
                  0.3671792,1,0.5884902,0.4380218,0.3148344,0.2399497,0.3783691,
                  0.3260069, 0.5884902, 1,0.474305,0.3441412,0.2264623,0.3448708,
                  0.2499102,0.4380218, 0.474305, 1,0.5431077,0.1560136,0.2078255,
                  0.2410074 ,0.3148344, 0.3441412, 0.5431077,1, 0.1661684,0.1880403,
                  0.2586295,0.2399497, 0.2264623,0.1560136, 0.1661684, 1,0.2315498,
                  0.1722615, 0.3783691, 0.3448708, 0.2078255, 0.1880403, 0.2315498, 1))

执行以下绘图代码后仅得到空白图：

ggplot(Nmi_frame, aes(x = mesure1, y = mesure2, fill = nmi_value),
        geom_tile(color = "white", aes(fill = Nmi_frame$nmi),
          scale_fill_gradient(low = "blue", high = "red")))

已尝试将前两列转为因子，但结果仍无改善。请问该如何解决此问题？是否为ggplot的bug？

问题原因与解决方法 #

这不是ggplot的bug，完全是代码语法错误导致的：

• 图层参数位置错误：geom_tile和scale_fill_gradient是独立的图层/标度函数，不能作为ggplot()的参数传入，必须用+符号连接在ggplot()之后，这是ggplot2的核心语法规则。
• 映射变量错误：aes(fill = Nmi_frame$nmi)中，Nmi_frame$nmi不存在，你的数据框里对应的列名是nmi_value；而且在ggplot的映射中不需要加数据框前缀，直接用列名即可。
• 冗余映射：你已经在ggplot()的全局aes里指定了fill = nmi_value，不需要在geom_tile里重复指定fill映射。

正确代码示例 #

library(ggplot2)

ggplot(Nmi_frame, aes(x = mesure1, y = mesure2, fill = nmi_value)) +
  geom_tile(color = "white") +
  scale_fill_gradient(low = "blue", high = "red") +
  # 可选：添加数值标签，让热力图更直观
  geom_text(aes(label = round(nmi_value, 2)), color = "white", size = 3) +
  labs(title = "不同聚类算法的归一化互信息热力图",
       x = "聚类算法", y = "聚类算法", fill = "NMI值") +
  theme_minimal()

代码说明 #

1. 用+连接各个组件：ggplot()定义基础数据和映射，后续用+添加geom_tile（绘制色块）、scale_fill_gradient（设置颜色渐变）等。
2. 全局映射复用：geom_tile直接继承ggplot()里的x、y、fill映射，只需要补充边框颜色参数。
3. 可选优化：添加geom_text显示NMI数值，用labs设置标题和标签，theme_minimal让图表更简洁美观。

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