Colour coding data points-CodePudding

I have some data for which I have performed manual K-means clustering using Euclidean distance. I have the final centroids and the clustered points. I have plotted this to a graph however, I am unsure how I could go about colour coding the different points depending on the cluster they belong to.

This is how the data has been clustered and I would like to colour the points depending on the colour they are in.

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)

Data

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))

Current implementation for the graph

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)

Current produced graph

CodePudding user response：

You will need to determine for each point which cluster they belong to and add this information to the plotting dataframe. Below is one approach:

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)

CodePudding user response：

You need to add the cluster information:

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

Now your plot can be done with:

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)