r - 如何在 fviz_silhouette 中将颜色更改为除簇号以外的变量

Question

我在R中使用 package factoextra来生成轮廓图。目前，剪影会通过聚类自动为您的图形着色。我想用我定义为 x 的另一个变量着色。我已经尝试将填充和颜色更改为站点变量，但似乎没有任何效果。我也尝试过使用and 。我认为关键在于另一个用户的源代码指出Sitescale_color_manualscale_fill_discretescale_fill_discrete

mapping <- aes_string(x = "name", y = "sil_width", color = "cluster", fill = "cluster")

我基本上需要更改color="cluster"为colour= "x".
我已将最终的绘图代码恢复为最基本的形式。

pamspec <- pam(spec, 3, keep.diss = TRUE) 
plot(pamspec)

spec <- cbind(pamspec$clustering)
autoplot(pam(spec,3), frame=TRUE, frame.type = "norm")

pamspec$site <- spec$Site
x <- pamspec$site
fviz_silhouette(pamspec, label=TRUE) + theme_classic()

标签： rggplot2cluster-computingsilhouette

Answer 1

fviz_silhouette我建议使用附加var.col输入参数的以下修改版本。

myfviz_silhouette <- function (sil.obj, var.col, label = FALSE, print.summary = TRUE, ...) {
    if (inherits(sil.obj, c("eclust", "hcut", "pam", "clara", 
        "fanny"))) {
        df <- as.data.frame(sil.obj$silinfo$widths, stringsAsFactors = TRUE)
    }
    else if (inherits(sil.obj, "silhouette")) 
        df <- as.data.frame(sil.obj[, 1:3], stringsAsFactors = TRUE)
    else stop("Don't support an oject of class ", class(sil.obj))
    df <- df[order(df$cluster, -df$sil_width), ]
    if (!is.null(rownames(df))) 
        df$name <- factor(rownames(df), levels = rownames(df))
    else df$name <- as.factor(1:nrow(df))
    df$cluster <- as.factor(df$cluster)
    df$var_col <- var.col
    mapping <- aes_string(x = "name", y = "sil_width", color = "var_col", 
        fill = "var_col")
    p <- ggplot(df, mapping) + geom_bar(stat = "identity") + 
        labs(y = "Silhouette width Si", x = "", title = paste0("Clusters silhouette plot ", 
            "\n Average silhouette width: ", round(mean(df$sil_width), 
                2))) + ggplot2::ylim(c(NA, 1)) + geom_hline(yintercept = mean(df$sil_width), 
        linetype = "dashed", color = "red")
    p <- ggpubr::ggpar(p, ...)
    if (!label) 
        p <- p + theme(axis.text.x = element_blank(), axis.ticks.x = element_blank())
    else if (label) 
        p <- p + theme(axis.text.x = element_text(angle = 45))
    ave <- tapply(df$sil_width, df$cluster, mean)
    n <- tapply(df$cluster, df$cluster, length)
    sil.sum <- data.frame(cluster = names(ave), size = n, ave.sil.width = round(ave, 
        2), stringsAsFactors = TRUE)
    if (print.summary) 
        print(sil.sum)
    p
}

以下是它的使用示例：

library(factoextra)
library(cluster)
pamspec <- pam(iris[,-5], 3, keep.diss = TRUE) 

color_var <- iris$Species
myfviz_silhouette(pamspec, color_var, label=TRUE) +
   theme_classic()