Tutorial 27: K-Means Clustering · Sharda Learning Center

Level: Intermediate · Part of: ML Learning Roadmap

← Back to roadmap

Case Study: Retail Customer Segments

Scenario

A retailer wants to group customers by purchase behavior for targeted marketing — without labeled segments.

Learning Objectives

Apply KMeans clustering
Choose k with elbow method and silhouette score
Interpret cluster centroids
Scale features before clustering

Prerequisites

Tutorial 05
Python 3.9+, scikit-learn, NumPy, pandas, matplotlib

Dataset

Synthetic customer purchase data

Hands-On Solution

Copy and run the complete script below:

import numpy as np
import matplotlib.pyplot as plt
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score

np.random.seed(42)
n = 300
cluster1 = np.random.randn(n // 3, 2) * 0.5 + [2, 8]
cluster2 = np.random.randn(n // 3, 2) * 0.5 + [8, 8]
cluster3 = np.random.randn(n - 2 * (n // 3), 2) * 0.5 + [5, 2]
X = np.vstack([cluster1, cluster2, cluster3])

X_scaled = StandardScaler().fit_transform(X)

inertias, silhouettes = [], []
K_range = range(2, 8)
for k in K_range:
    km = KMeans(n_clusters=k, random_state=42, n_init=10)
    labels = km.fit_predict(X_scaled)
    inertias.append(km.inertia_)
    silhouettes.append(silhouette_score(X_scaled, labels))

print("Silhouette scores:", dict(zip(K_range, [f"{s:.3f}" for s in silhouettes])))

km = KMeans(n_clusters=3, random_state=42, n_init=10)
labels = km.fit_predict(X_scaled)

plt.figure(figsize=(7, 5))
plt.scatter(X_scaled[:, 0], X_scaled[:, 1], c=labels, cmap="viridis", alpha=0.7)
plt.scatter(km.cluster_centers_[:, 0], km.cluster_centers_[:, 1],
            marker="X", s=200, c="red", label="Centroids")
plt.legend()
plt.savefig("kmeans_segments.png", dpi=120)
print("Saved kmeans_segments.png")

Expected Output

When you run the script, you should see evaluation metrics printed to the console. Some tutorials also save .png plot files in the current directory.

Exercises

Plot the elbow curve (k vs inertia).
Try n_init=1 vs n_init=10. Why does sklearn warn?
Apply KMeans to load_iris() without labels. Compare clusters to true species.

Key Takeaways

K-Means partitions data into k spherical clusters
Always scale features — clustering is distance-based
Silhouette score helps choose k when ground truth is unknown

← Tutorial 26: PCA for Dimensionality Reduction

Tutorial 28: Hierarchical Clustering →

Part of the ML Learning Roadmap — Hands-On with scikit-learn