Welcome to k-means clustering! Remember unsupervised learning from our first unit? k-means is a perfect example - it finds hidden groups in data without any labels.

Instead of predicting outcomes, k-means discovers natural clusters by grouping similar data points together.

Engagement Message

Can you share one real-world example where you'd want to segment customers without any labels.

k-means works by finding centroids—the center points of each cluster. Think of centroids as the "average location" of all the points in a group.

The algorithm starts with random centroids, then shifts them to better reflect the positions of their assigned points.

Engagement Message

Does this make sense?

Let's look at a simple example to see how k-means works in practice. Here's our simple dataset of 6 points on a 2D plane:

Notice how some points are close together? k-means will discover these natural groups.

Engagement Message

Does this make sense?

Let's start with k=2 (two clusters). We randomly place our initial centroids at:

• Centroid 1: (3, 3)
• Centroid 2: (6, 6)

Now we assign each point to its nearest centroid using Euclidean distance.

Engagement Message

Can you type the letters of the points you think will be closer to Centroid 1?