Clustering.md

K-means Clustering

Jesse Cambon 24 November, 2019

Demonstrate K-means clustering

References

• https://uc-r.github.io/kmeans_clustering
• https://cran.r-project.org/web/packages/broom/vignettes/kmeans.html

Setup

library(tidyverse)
library(knitr)
library(kableExtra)
library(broom)
library(factoextra)
library(wesanderson)

Cluster Data

library(fueleconomy)
my_vehicles <- vehicles %>% filter(year == 2015) %>%
  drop_na(cyl,displ,cty)

vehicles_sel <- my_vehicles %>%
  select(cty,cyl,displ)

# Scale variables for clustering
vehicles_sel_scaled <- vehicles_sel %>%
  mutate_all(scale)

# Try different numbers of clusters
clust3 <- kmeans(vehicles_sel_scaled, centers = 3)
clust5 <- kmeans(vehicles_sel_scaled, centers = 5)
clust7 <- kmeans(vehicles_sel_scaled, centers = 7) 
clust10 <- kmeans(vehicles_sel_scaled, centers = 10)

View Results

combine_results <- augment(clust3, my_vehicles)  

combine_summ <- combine_results %>% group_by(.cluster) %>% 
  summarize(num_vehicles=n(),
            mean_cty=mean(cty),
            min_cty=min(cty),
            max_cty=max(cty),
            mean_displ=mean(displ),
            mean_cyl=mean(cyl)) %>%
  arrange(desc(mean_cty))

kable(combine_summ,format='markdown',digits=2) %>%
  kable_styling(bootstrap_options = c("striped",'border'))

.cluster	num_vehicles	mean_cty	min_cty	max_cty	mean_displ	mean_cyl
3	86	25.05	20	33	1.87	4.05
2	55	18.71	16	24	3.12	5.78
1	63	15.19	11	20	5.12	8.29

Visualize

fviz_cluster(clust3,data=vehicles_sel,repel=F,ggtheme=theme_bw())

ggplot(aes(x=cyl,y=cty,color=.cluster),data=combine_results) + 
  geom_jitter() +
  theme_bw() +
  theme(legend.position='top') +
  scale_color_manual(values=wes_palette('Darjeeling1')) +
  guides(color = guide_legend(title='Cluster',override.aes = list(size=2.5))) +
  xlab('Cylinders (cyl)') +
  ylab('City Fuel Economy (cty)')