R语言PAM聚类与t-SNE可视化报错：mutate不适用于factor类对象

一、问题根源 #

报错Error in UseMethod("mutate") : no applicable method for 'mutate' applied to an object of class "factor"说明执行mutate前的对象是factor类型，而dplyr::mutate仅支持数据框/tibble对象。

二、分步排查与修复 #

1. 检查并修正数据列类型

先确认数据集df2的数值列不是factor类型：

str(df2)

如果health_expen、gdp_pc等列显示为Factor，转换为数值型：

library(dplyr)
df2 <- df2 %>% 
  mutate(across(c(health_expen, gdp_pc, life_exp, healthsoc_empl), as.numeric))

2. 明确调用dplyr的mutate函数

避免与其他包（如plyr）的同名函数冲突，所有mutate前加上dplyr::前缀：

# 聚类分析部分修正
pam_results <- df2 %>%
  dplyr::select(-country) %>%
  dplyr::mutate(cluster = pam_fit$clustering) %>%
  dplyr::group_by(cluster) %>%
  dplyr::do(the_summary = summary(.))

# t-SNE可视化部分修正
tsne_data <- tsne_obj$Y %>%
  data.frame() %>%
  setNames(c("X", "Y")) %>%
  dplyr::mutate(cluster = factor(pam_fit$clustering), name = df2$country)

3. 修正t-SNE的维度与参数设置

你设置了dims=3（3维），但后续只命名了2列（X、Y），建议统一维度为2（适配2D可视化），同时调整perplexity值（样本数为6时，perplexity建议设为2-5）：

tsne_obj <- Rtsne(gower_dist, is_distance = TRUE, dims = 2, perplexity = 2)

4. 验证Gower距离矩阵的正确性

确保gower_dist是标准的距离矩阵（class为dist），用cluster::daisy计算：

library(cluster)
gower_dist <- daisy(df2[, -1], metric = "gower")
class(gower_dist) # 输出应为"dist"

三、完整可运行代码示例 #

library(cluster)
library(dplyr)
library(Rtsne)
library(ggplot2)

# 构造数据集
df2 <- tibble(
  country = c("Australia", "Austria", "Belgium", "Canada", "Chile", "Colombia"),
  health_expen = c(5130., 5624., 5353., 5190., 2297., 1293.),
  gdp_pc = c(52669., 58076., 54278., 49299., 25189., 15662.),
  life_exp = c(82.9, 82, 82.1, 82.3, 80.6, 76.6),
  healthsoc_empl = c(68, 53.0, 56.4, 59.9, 26.5, 18.4)
)

# 计算Gower距离
gower_dist <- daisy(df2[, -1], metric = "gower")

# PAM聚类
pam_fit <- pam(gower_dist, diss = TRUE, k = 2)

# 聚类结果分析
pam_results <- df2 %>%
  dplyr::select(-country) %>%
  dplyr::mutate(cluster = pam_fit$clustering) %>%
  dplyr::group_by(cluster) %>%
  dplyr::do(the_summary = summary(.))
pam_results$the_summary

# 查看medoids
df2[pam_fit$medoids, ]

# t-SNE可视化
tsne_obj <- Rtsne(gower_dist, is_distance = TRUE, dims = 2, perplexity = 2)
tsne_data <- tsne_obj$Y %>%
  data.frame() %>%
  setNames(c("X", "Y")) %>%
  dplyr::mutate(cluster = factor(pam_fit$clustering), name = df2$country)

ggplot(tsne_data, aes(x = X, y = Y)) +
  geom_point(aes(color = cluster), size = 3) +
  geom_text(aes(label = name), vjust = -1)

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