KNN, Hierarchical, and EM Clustering Implementations in R

This package was completed as part of STAT 431 Advanced Programming in R at Cal Poly

library(clust431)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(ggplot2)

K Nearest Neighbors

First the classic of example of predicting iris species using the petal and sepal lengths. We also compute the proportion of variation explained by the grouping. Each one of my implementations is compared to the base R equivalent.

kmodel <- kmeans(iris[,1:4], 3)
kmodel$betweenss / kmodel$totss
#> [1] 0.8842753
table(iris$Species, kmodel$cluster)
#>             
#>               1  2  3
#>   setosa      0  0 50
#>   versicolor 48  2  0
#>   virginica  14 36  0

mymodel <- k_means(iris[,1:4], 3)
mymodel$ssbetween / mymodel$sstotal
#> [1] 0.7851997
table(iris$Species, mymodel$assign)
#>             
#>               1  2  3
#>   setosa     18 32  0
#>   versicolor  4  0 46
#>   virginica   0  0 50

Now to predict the number of cylinders in a car using othe rmetrics (mtcars). This time we will use the first two principle components instead of the original variables.

mttrain <- mtcars %>%
  select(mpg, disp, hp, drat, wt, qsec)

mtcarspc <- data.frame((princomp(mttrain)$scores[,1:2]))

kmodel <- kmeans(mtcarspc, 3)
kmodel$betweenss / kmodel$totss
#> [1] 0.8539862
table(mtcars$cyl, kmodel$cluster)
#>    
#>      1  2  3
#>   4 11  0  0
#>   6  5  0  2
#>   8  0  9  5

mymodel <- k_means(mttrain, 3, T)
mymodel$ssbetween / mymodel$sstotal
#> [1] 0.8343965
table(mtcars$cyl, mymodel$assign)
#>    
#>      1  2  3
#>   4 11  0  0
#>   6  6  0  1
#>   8  0  4 10

Hierarchical Clustering

mydata <- iris[,1:4]
groupstrs <- hier_clust(mydata, 3)
print(groupstrs)
#> [1] "16, 15, 34, 33, 19, 6, 17, 49, 11, 47, 22, 20, 45, 37, 32, 21, 44, 27, 24, 36, 50, 40, 8, 18, 1, 29, 28, 41, 38, 5, 25, 12, 31, 30, 26, 35, 10, 13, 46, 2, 7, 3, 48, 4, 42, 23, 14, 43, 39, 9"                                                                                                                                                                                                   
#> [2] "61, 99, 94, 58, 107, 85, 67, 91, 56, 89, 97, 96, 100, 95, 68, 93, 83, 74, 79, 92, 64, 62, 72, 98, 75, 63, 80, 65, 60, 90, 54, 70, 82, 81, 76, 66, 59, 55, 77, 87, 53, 51, 86, 57, 52, 88, 69, 120, 147, 127, 124, 73, 134, 84, 150, 71, 139, 128, 115, 122, 114, 143, 102, 78, 148, 111, 105, 133, 129, 112, 104, 138, 117, 116, 149, 137, 125, 144, 121, 145, 141, 140, 113, 146, 142, 135, 109"
#> [3] "101, 119, 123, 106, 136, 131, 108, 103, 130, 126, 110, 132, 118"

EM Algorithm

dat <- iris[,1:4]

iris_clust <- em_clust(dat, 3)

table(iris_clust$group, iris$Species)
#>    
#>     setosa versicolor virginica
#>   1     50          0         0
#>   2      0          3        46
#>   3      0         47         4

plotone <- ggplot(data = iris, aes(Sepal.Length, Sepal.Width, colour = Species)) + 
    geom_point(shape = iris_clust$group) + 
    geom_point(data = iris_clust$mu, aes(iris_clust$mu[,1], iris_clust$mu[,2]), color = "blue", size = 3)

plottwo <- ggplot(data = iris, aes(Petal.Length, Petal.Width, colour = Species)) + 
    geom_point(shape = iris_clust$group) + 
    geom_point(data = iris_clust$mu, aes(iris_clust$mu[,3], iris_clust$mu[,4]), color = "blue", size = 3)
    
plotone

plottwo