CLARANS 2D Clustering on Synthetic Blobs

This example demonstrates how CLARANS identifies clusters and true medoids (actual data points) on a 2D Gaussian blobs dataset.

# Authors: Ngọc Thiện Nguyễn <thiennguyen03001@gmail.com>
# License: MIT

import matplotlib.pyplot as plt
from sklearn.datasets import make_blobs
from clarans import CLARANS

COLORS = ["#2b5c8f", "#d95f02", "#7570b3", "#1b9e77"]


X, _ = make_blobs(n_samples=500, centers=4, cluster_std=0.60, random_state=42)
model = CLARANS(n_clusters=4, num_local=5, random_state=42)
model.fit(X)

fig, ax = plt.subplots(figsize=(6.0, 4.5))
for k in range(4):
    mask = model.labels_ == k
    ax.scatter(X[mask, 0], X[mask, 1], c=COLORS[k], s=25, alpha=0.65, edgecolors="none")

centers = model.cluster_centers_
ax.scatter(
    centers[:, 0],
    centers[:, 1],
    marker="x",
    s=90,
    c="black",
    linewidths=1.8,
    label="Medoids",
)

ax.set_title("CLARANS on Synthetic Blobs ($k=4$)", fontsize=11, pad=8)
ax.set_xticks([])
ax.set_yticks([])
for spine in ax.spines.values():
    spine.set_color("#cccccc")
    spine.set_linewidth(0.8)

ax.legend(loc="upper right", frameon=False, fontsize=9)
plt.tight_layout()
plt.show()