How to Use kmeans in R: Simple Guide with Example

Use the kmeans() function in R by providing a numeric dataset and the number of clusters you want. It returns cluster centers and assignments, which help group similar data points.

Syntax

The basic syntax of kmeans() is:

• x: numeric data matrix or data frame.
• centers: number of clusters or initial cluster centers.
• nstart: how many random sets of centers to try (optional, improves results).
• iter.max: max iterations allowed (optional).

r

kmeans(x, centers, nstart = 1, iter.max = 10)

Example

This example shows how to cluster 2D points into 3 groups using kmeans(). It prints cluster centers and the cluster assignment for each point.

r

set.seed(123)
data <- matrix(rnorm(30), ncol=2)
kmeans_result <- kmeans(data, centers=3, nstart=20)
print(kmeans_result$centers)
print(kmeans_result$cluster)

Output

           [,1]       [,2]
1  0.7150650  0.1216750
2 -0.6879627 -0.5757814
3 -0.3053884  0.0918203
[1] 3 1 3 2 2 2 1 1 3 2 3 1 1 3 2 3 1 2 2 1 3 2 1 3 2 1 3 2 1 3

Common Pitfalls

Common mistakes include:

• Using non-numeric data, which causes errors.
• Not setting nstart, which can lead to poor clustering due to bad initial centers.
• Choosing too many or too few clusters without checking results.

Always scale data if features have different units.

r

## Wrong: Using character data
# kmeans(c('a','b','c'), centers=2)

## Right: Use numeric data and set nstart
set.seed(1)
data <- matrix(rnorm(20), ncol=2)
kmeans(data, centers=2, nstart=10)

Quick Reference

Tips for using kmeans() effectively:

• Always set nstart to a value like 10 or 20 for better results.
• Scale your data if variables have different scales.
• Check cluster centers and sizes to understand results.
• Visualize clusters when possible.

Key Takeaways

Use kmeans() with numeric data and specify the number of clusters.

Set nstart to try multiple initial centers for better clustering.

Avoid non-numeric data and scale features if they differ in units.

Check cluster centers and assignments to interpret results.

Visualize clusters to better understand grouping.