kluster: Automatically cluster 1-dimensional continuous data.

In sift: Intelligently Peruse Data

Description

Automatically cluster 1-dimensional continuous data.

Usage

kluster(x, bw = "SJ", fixed = FALSE)

Arguments

x	Vector to be clustered. Must contain at least 1 non-missing value.
bw	kernel bandwidth. Default "SJ" should suffice more application, however you can supply a custom numeric value. See ?stats::density for more information.
fixed	logical; if TRUE, performs simple 1-dimensional clustering without kernel density estimation. default FALSE.

Value

An integer vector identifying the cluster corresponding to each element in x.

Examples

# Below vector clearly has 2 groups.
# kluster will identify these groups using kernel density estimation.
kluster(c(0.1, 0.2, 1))

# kluster shines in cases where manually assigning groups via "eyeballing" is impractical.
# Suppose we obtained vector 'x' without knowing how it was generated.
set.seed(1)
nodes <- runif(10, min = 0, max = 100)
x <- lapply(nodes, function(x) rnorm(10, mean = x, sd = 0.1))
x <- unlist(x)

kluster(x) # kluster reveals the natural grouping

kluster(x, bw = 10) # adjust bandwidth depending on application

# Example with faithful dataset
faithful$k <- kluster(faithful$eruptions)
library(ggplot2)
ggplot(faithful, aes(eruptions)) +
  geom_density() +
  geom_rug(aes(color = factor(k))) +
  theme_minimal() +
  scale_color_discrete(name = "k")

sift documentation built on July 5, 2021, 5:08 p.m.