使用fviz_cluster可视化CLARA聚类结果时如何突出中心点medoid

方案1：调整原生参数突出中心点 #

fviz_cluster本身提供了中心点样式的自定义参数，直接设置即可让中心点盖在普通数据点上方，不会被密集的样本淹没：

library(factoextra)
library(ggplot2)

p <- fviz_cluster(clara.res,
         palette = c("#004c6d",
                     "#00ffff",
                     "#00a1c1",
                     "#6efa75",
                     "#78ab63",
                     "#cc0089",
                     "#ffc334",
                     "#ff9509",
                     "#ffb6de",
                     "#00cfe3"
                      ),
         ellipse.type = "t",
         geom = "point",
         show.clust.cent = TRUE,
         # 新增中心点自定义参数
         cent.point.size = 10, # 中心点尺寸远大于普通点的0.5
         cent.point.shape = 23, # 菱形带填充的形状，和普通圆点区分
         cent.point.color = "black", # 中心点边框颜色
         cent.point.fill = "red", # 中心点填充色，和聚类颜色区分开
         repel = TRUE,
         pointsize = 0.5,
         alpha = 0.1, # 普通点调低透明度，避免糊成一团遮挡中心点
         ggtheme = theme_classic()
)
p

方案2：隐藏所有普通数据点仅保留中心点 #

只需要给普通点设置完全透明即可，聚类椭圆、坐标轴等元素都可以正常保留：

p_hide_points <- fviz_cluster(clara.res,
         palette = c("#004c6d",
                     "#00ffff",
                     "#00a1c1",
                     "#6efa75",
                     "#78ab63",
                     "#cc0089",
                     "#ffc334",
                     "#ff9509",
                     "#ffb6de",
                     "#00cfe3"
                      ),
         ellipse.type = "t",
         geom = "point",
         show.clust.cent = TRUE,
         cent.point.size = 10,
         cent.point.shape = 23,
         cent.point.color = "black",
         cent.point.fill = "red",
         repel = TRUE,
         pointsize = 0.5,
         alpha = 0, # 普通点完全透明，相当于隐藏
         ggtheme = theme_classic()
)
p_hide_points

方案3：手动添加你已获取的medoid #

fviz_cluster返回的是标准ggplot2对象，你可以直接叠加自定义的点层，完全可控中心点的样式，优先级最高不会被遮挡。
假设你已将medoid对应的行号存储在向量medoid_rows中，操作如下：

# 先生成基础绘图对象，关闭默认中心点展示
p_base <- fviz_cluster(clara.res,
         palette = c("#004c6d",
                     "#00ffff",
                     "#00a1c1",
                     "#6efa75",
                     "#78ab63",
                     "#cc0089",
                     "#ffc334",
                     "#ff9509",
                     "#ffb6de",
                     "#00cfe3"
                      ),
         ellipse.type = "t",
         geom = "point",
         show.clust.cent = FALSE,
         repel = TRUE,
         pointsize = 0.5,
         alpha = 0.1,
         ggtheme = theme_classic()
)

# 提取medoid对应的降维坐标
medoid_data <- p_base$data[medoid_rows, ]

# 手动叠加中心点层，显示在最上层
p_custom_medoid <- p_base +
  geom_point(data = medoid_data, aes(x = x, y = y),
             shape = 23, 
             fill = "red",
             color = "black",
             size = 10,
             alpha = 1)

p_custom_medoid

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