Clustering

Dendrograms

What cars are related to one another?

1. Calculate the “distance” between each pair of cars

Dists <- dist(mtcars)

2. Agglomerate them recursively

Dendrogram <- hclust(Dists)
ddata <- dendro_data(Dendrogram)
ggdendrogram(Dendrogram, rotate = TRUE) +
  geom_text(data = ddata$labels, aes(x=x, y=y, label = label), hjust = 0, vjust = 1) +
  theme_void()

Irises

Flowers <- iris %>% mutate(ID = paste(Species, 1:n())) 
rownames(Flowers) <- Flowers$ID
Dists <- dist(dplyr::select(Flowers, -ID, -Species))
Dendrogram <- hclust(Dists)
ddata <- dendro_data(Dendrogram)
ddata$labels$label <- Flowers$Species
ggdendrogram(Dendrogram, rotate = TRUE, labels = TRUE) +
  geom_text(data = ddata$labels, aes(x = x, y = y, label = nchar(as.character(label)), color = as.factor(nchar(as.character(label))))) +
  theme_void()

K-means

clusts <- kmeans(dplyr::select(iris, -Species), centers = 5)
iris$cluster <- as.factor(clusts$cluster)
iris %>% 
  ggplot(aes(x = Petal.Length, y = Petal.Width, color = cluster, shape = Species)) + geom_point()

Countries

Locations <- dplyr::select(WorldCities, latitude, longitude)
city_clusts <- kmeans(Locations, centers = 6)
Locations$clust <- as.factor(city_clusts$cluster)
Locations %>% 
  ggplot(aes(x = longitude, y = latitude, color = clust)) +
  geom_point()

```