如何让自定义K均值聚类PC投影函数返回ggplot对象？

Got it, let's tweak your function to return a ggplot object properly. Here's a complete, modified version with explanations of key changes:

plot_kmeans_pc <- function(feature_matrix, k, pc) {
  # Check for ggplot2 dependency and throw clear error if missing
  if (!requireNamespace("ggplot2", quietly = TRUE)) {
    stop("The ggplot2 package is required to run this function. Install it first with install.packages('ggplot2')")
  }
  
  # Get input matrix name for plot title
  matrix_name <- deparse(substitute(feature_matrix))
  
  # Run K-means clustering
  pclusters <- kmeans(feature_matrix, k, nstart = 100, iter.max = 100)
  groups <- pclusters$cluster
  
  # Project data onto first two principal components
  projected <- predict(pc, newdata = feature_matrix)[, 1:2]
  
  # Build a clean data frame for ggplot
  projected_df <- as.data.frame(projected)
  projected_df$cluster <- factor(groups) # Treat cluster as categorical for better color handling
  
  # Create and configure the ggplot object
  cluster_plot <- ggplot2::ggplot(projected_df, ggplot2::aes(x = PC1, y = PC2, color = cluster)) +
    ggplot2::geom_point(alpha = 0.7) + # Add transparency for dense datasets
    ggplot2::labs(
      title = paste("K-Means Clustering (k =", k, ") on", matrix_name),
      x = "Principal Component 1",
      y = "Principal Component 2",
      color = "Cluster Label"
    ) +
    ggplot2::theme_minimal()
  
  # Return the ggplot object instead of printing it directly
  return(cluster_plot)
}

Key Changes Explained #

• Dependency Check: Added a check to ensure ggplot2 is installed, with a user-friendly error message if it's missing.
• Data Frame Completion: Properly combined projected PC values with cluster labels, converting cluster to a factor so ggplot treats it as a categorical variable (improves color scale behavior).
• Explicit ggplot2 Calls: Used ggplot2:: prefixes to avoid namespace conflicts if ggplot2 isn't loaded in the global environment.
• Return the ggplot Object: Instead of rendering the plot immediately, we assign it to a variable and return it. This lets you store the plot, modify it later (e.g., add annotations, adjust themes), or save it with ggsave().

Example Usage #

# 1. Generate sample data and train a PCA model
set.seed(123)
sample_data <- matrix(rnorm(200*6), ncol = 6)
pca_model <- prcomp(sample_data, scale. = TRUE)

# 2. Get the ggplot object
my_plot <- plot_kmeans_pc(sample_data, k = 4, pc = pca_model)

# 3. Display the plot
print(my_plot)

# 4. Modify the plot further if needed
my_plot + ggplot2::ggtitle("Custom Cluster Plot Title") + ggplot2::theme(plot.title = ggplot2::element_text(hjust = 0.5))

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