In this article, I show the implementation of K-mean algorithm using open source free AI tools. You can use ChatGPT, Claude AI, Blackbox AI, Perplexity AI, etc
SIMPLE K-MEANS IMPLEMENTATION FROM SCRATCH¶
In [ ]:
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import make_blobs
In [ ]:
1. CREATE SIMPLE DATASET¶
Prompt: Generate code to simulate a synthetic dataset for a clustering task. The implementation must follow these requirements:
- Generate a dataset with exactly 15 samples. It should be distributed across 3 distinct cluster centers.
- Standard deviation of the clusters should be 1.0.
- Fix the random seed to ensure the dataset is reproducible.
- Store the generated data points in a variable named ‘X’ and their true cluster labels in a variable named ‘y’.
- Plot the points
Write this cleanly using python and numpy or any package that is available in data science.
In [ ]:
X, y = make_blobs(
n_samples=15, # Experiment with 30, 90, 300
centers=3,
cluster_std=1.0,
random_state=42
)
# Plot original data
plt.figure(figsize=(6, 5))
plt.scatter(X[:, 0], X[:, 1])
plt.title("Original Data")
plt.show()
In [ ]:
X
Out[ ]:
array([[-0.92998481, 9.78172086],
[-2.97261532, 8.54855637],
[ 6.10552761, 1.74739338],
[-8.03062077, -6.50441157],
[ 4.70740704, 0.5484215 ],
[-6.05708228, -8.10095324],
[ 3.73185476, 0.56086598],
[-4.23411546, 8.4519986 ],
[-7.48026588, -7.17180334],
[-2.26723535, 7.10100588],
[ 4.09549611, 2.08409227],
[-7.4813338 , -5.02783141],
[ 3.62704772, 2.28741702],
[-6.89312442, -7.93782052],
[-2.97867201, 9.55684617]])
In [ ]:
In [ ]:
In [ ]:
2. SIMPLE K-MEANS FUNCTION¶
Prompt: Implement following psuedocode
FUNCTION kmeans(X, k=3, max_iterations=100):
STEP 1: INITIALIZATION
- Select 'k' unique random items from dataset X.
- Store these 'k' items in a list named 'centroids'.
LOOP: Repeat for a maximum of 'max_iterations':
STEP 2: CLUSTER ASSIGNMENT
- Create an empty list named 'clusters'.
- FOR each 'point' in X:
- Create an empty list named 'distances'.
- FOR each 'centroid' in 'centroids':
- Calculate the Euclidean distance between 'point' and 'centroid'.
- Append this distance to 'distances'.
- Find the index of the smallest value in 'distances'.
- Append this index to 'clusters'.
STEP 3: CENTROID UPDATE
- Create an empty list named 'new_centroids'.
- FOR each cluster index 'i' from 0 to k-1:
- Create an empty list named 'cluster_points'.
- FOR each index 'j' from 0 to length of X - 1:
- IF clusters[j] equals 'i':
- Append X[j] to 'cluster_points'.
- Calculate the average (mean) position of all vectors in 'cluster_points'.
- Append this average vector to 'new_centroids'.
STEP 4: CONVERGENCE CHECK
- IF 'centroids' matches 'new_centroids' exactly in values and position:
- Break out of the loop.
- Set 'centroids' equal to 'new_centroids'.
RETURN clusters, centroidsIn [ ]:
def kmeans(X, k=3, max_iterations=100):
# Step 1: Randomly choose initial centroids
random_indices = np.random.choice(len(X), k, replace=False)
centroids = X[random_indices]
# Repeat until convergence
for iteration in range(max_iterations):
# ====================================
# Step 2: Assign clusters
# ====================================
clusters = []
for point in X:
# Distance from point to each centroid
distances = []
for centroid in centroids:
distance = np.linalg.norm(point - centroid)
distances.append(distance)
# Choose nearest centroid
cluster = np.argmin(distances)
clusters.append(cluster)
clusters = np.array(clusters)
# ====================================
# Step 3: Update centroids
# ====================================
new_centroids = []
for i in range(k):
# Points belonging to cluster i
cluster_points = X[clusters == i]
# Mean of cluster points
centroid = cluster_points.mean(axis=0)
new_centroids.append(centroid)
new_centroids = np.array(new_centroids)
# ====================================
# Step 4: Stop if centroids don't move
# ====================================
if np.all(centroids == new_centroids):
break
centroids = new_centroids
return clusters, centroids
In [ ]:
3. RUN K-MEANS¶
In [ ]:
clusters, centroids = kmeans(X, k=3)
print("Final Centroids:\n")
print(centroids)
print("\n\nClusters:\n")
print(clusters)
Final Centroids: [[-7.18848543 -6.94856402] [-2.67652459 8.68802558] [ 4.45346665 1.44563803]] Clusters: [1 1 2 0 2 0 2 1 0 1 2 0 2 0 1]
In [ ]:
4. VISUALIZE CLUSTERS¶
Prompt: Write a Python function using matplotlib that takes two inputs, ‘X’ and ‘centroids’, and plots them as a static 2D scatter plot.
Requirements:
- Plot ‘X’ as blue circles and ‘centroids’ as red ‘X’ markers.
- Include a title, axis labels, gridlines, and a legend.
- Display the final plot.
In [ ]:
plt.figure(figsize=(7, 5))
# Plot clustered points
plt.scatter(
X[:, 0],
X[:, 1],
c=clusters
)
# Plot centroids
plt.scatter(
centroids[:, 0],
centroids[:, 1],
s=300,
marker='X'
)
plt.title("K-Means Clustering From Scratch")
plt.show()
In [ ]:
Nice Introduction to using AI for implementation