颜色编码数据点

Question

我有一些数据，我已使用欧氏距离对其执行了手动 K 均值聚类。我有最终的质心和聚类点。我已经将其绘制成图表，但是我不确定如何根据它们所属的集群对不同的点进行颜色编码。

这就是数据聚类的方式，我想根据点的颜色为它们着色。

Cluster1 centroid location: 2, 3.5
points under Cluster1: (1,1), (1,6), (2,1), (4,6)

Centroid2 centroid location: 6.2, 8.8
points under Cluster2: (3,9), (3,10), (5,6), (8,9), (9,9), (9,10)

Centroid3 centroid location: 8.8, 2.4
points under Cluster3: (7,2), (8,1), (9,1), (10,3), (10,5)

数据

structure(list(Subject = 1:15, X = c(1L, 1L, 2L, 3L, 3L, 4L, 
5L, 7L, 8L, 8L, 9L, 9L, 9L, 10L, 10L), Y = c(1L, 6L, 1L, 9L, 
10L, 6L, 6L, 2L, 1L, 9L, 1L, 9L, 10L, 3L, 5L), E1 = c(1L, 1L, 
NA, NA, NA, 1L, 1L, 1L, 1L, 1L, NA, 1L, 1L, 1L, NA), E2 = c(NA, 
NA, 2L, 2L, 2L, NA, NA, NA, NA, NA, 2L, NA, NA, NA, 2L)), class = "data.frame", row.names = c(NA, 
-15L))

图的当前实现

library(tidyverse)
library(ggplot2)
data <- read.csv("data.csv")

data2 <- data[, -c(1)]
data2 <- data %>% filter(!if_all(c(E1, E2), is.na)) %>% mutate(E = ifelse(is.na(E1), E2, E1))

# Format centroid positions to be a 2D point with coordinates
x <- c(2,   6.2, 8.8)
y <- c(3.5, 8.8, 2.4)
coords = paste(x,y, sep=",")
df = data.frame(x,y)


ggplot(data2, aes(X, Y, shape = factor(E))) +
  geom_point(size = 4) +
  scale_shape_manual(values = c(8, 3), name = "E") +
  theme_bw() + 
  geom_point(df, mapping = aes(x, y), col = "blue", size = 3, inherit.aes = FALSE) +
  geom_label(df, mapping = aes(x + .5, y + 0.5, label = coords), inherit.aes = FALSE)

当前生成的图

Answer 1

您需要确定每个点属于哪个簇，并将此信息添加到绘图数据框中。以下是一种方法：

df %>% 
    # add group
    mutate(group = factor(row_number())) %>% 
    
    # create all combinations with data2 (rows for each point in data2 with each centroid)
    crossing(data2) %>% 
    
    # compute euclidean distance
    mutate(dist = (X-x)^2 + (Y-y)^2) %>% 
    
    # for each subject, filter for the closest centroid
    group_by(Subject) %>% 
    slice_min(dist) %>% 
    ungroup() %>% 
    
    # plot
    ggplot(aes(colour = group, shape = group)) + 
    geom_point(aes(X, Y), size = 3) +
    geom_point(aes(x, y), size = 5)

Answer 2

您需要添加集群信息：

data2$Cluster <- factor(c(1, 1, 1, 2, 2, 1, 2, 3, 3, 2, 3, 2, 2, 3, 3))
df$Cluster <- factor(1:3)

现在你的情节可以用：

ggplot(data2, aes(X, Y, shape = factor(is.na(E1)))) +
  geom_point(size = 4, aes(color = Cluster)) +
  scale_shape_manual(values = c(8, 3), name = "E", labels = 1:2) +
  theme_bw() + 
  geom_point(data = df, aes(x, y, color = Cluster), size = 4,
             inherit.aes = FALSE) +
  geom_label(data = df, aes(x + .5, y + 0.5, label = coords), 
             inherit.aes = FALSE)

颜色编码数据点

Colour coding data points

r

ggplot2

cluster-analysis