After completing this lesson, you will be able to:
Businesses collect enormous amounts of customer data from websites, mobile applications, online purchases, loyalty programs, social media, and customer relationship management (CRM) systems. Although this data contains valuable insights, organizations often do not know in advance how customers should be grouped.
This is where Clustering becomes valuable. Clustering is an Unsupervised Machine Learning technique that automatically groups similar data points based on their characteristics without requiring predefined labels.
Unlike supervised learning algorithms that predict known outcomes, clustering discovers hidden structures and natural groupings within data. These groups, known as clusters, help organizations understand customer behavior, identify market segments, detect anomalies, and develop personalized marketing strategies.
Today, clustering is widely used in retail, banking, healthcare, insurance, e-commerce, telecommunications, manufacturing, and digital marketing. Companies such as Amazon, Netflix, Spotify, and Google use clustering techniques to understand user behavior and deliver personalized experiences.
In this lesson, you will learn the fundamentals of clustering, understand how it differs from classification, explore the most common distance measures, and discover how businesses use clustering for customer segmentation.
Clustering is an Unsupervised Machine Learning technique that groups similar observations into clusters based on shared characteristics.
The objective is to maximize similarity among data points within the same cluster while maximizing differences between different clusters.
Unlike classification algorithms, clustering does not require predefined categories or target variables. Instead, it automatically discovers hidden patterns within the dataset.
For example, a retail company may have thousands of customers with different purchasing habits. Instead of manually defining customer categories, a clustering algorithm automatically groups customers with similar shopping behavior.
Common clustering applications include:
Although clustering and classification both group data, they solve different types of Machine Learning problems.
| Clustering | Classification |
|---|---|
| Unsupervised Learning | Supervised Learning |
| No target variable | Target variable is known |
| Discovers hidden groups | Predicts predefined classes |
| No labeled data | Requires labeled data |
| Customer Segmentation | Email Spam Detection |
| Market Segmentation | Loan Approval |
| Recommendation Systems | Customer Churn Prediction |
In simple terms, classification predicts “Which category does this belong to?”, whereas clustering discovers “What natural groups exist in this data?”.
Clustering belongs to the category of Unsupervised Learning because the algorithm learns without labeled training data.
There are no predefined answers or target variables. Instead, the algorithm analyzes similarities between observations and automatically forms groups.
For example, if an e-commerce company has customer information such as age, income, spending score, and purchase frequency, the clustering algorithm groups similar customers without being told which customer belongs to which segment.
This makes clustering particularly useful when businesses have large amounts of data but no predefined customer categories.
Effective clustering algorithms share several important characteristics.
Because clustering is exploratory, analysts often use it as one of the first steps in understanding a new dataset.
Most clustering algorithms determine similarity by measuring the distance between data points. Smaller distances indicate greater similarity, while larger distances indicate greater differences.
The choice of distance measure can significantly influence clustering results.
Euclidean Distance is the most commonly used distance metric in clustering algorithms such as K-Means.
It measures the straight-line distance between two points in multidimensional space.
For example, if two customers have similar ages and annual incomes, the Euclidean distance between them will be relatively small, increasing the likelihood that they belong to the same cluster.
Euclidean Distance works best when numerical variables have been standardized or normalized.
Manhattan Distance, also called City Block Distance, calculates the distance by moving along horizontal and vertical paths instead of a straight line.
This measure is particularly useful when movement follows a grid-like structure, such as city streets.
In data analytics, Manhattan Distance is often preferred when datasets contain high-dimensional features or when robustness to outliers is desired.
Cosine Similarity measures the angle between two vectors rather than the physical distance between them.
Instead of comparing magnitudes, it compares the direction of data points.
Cosine Similarity is widely used in:
For example, two customers who purchase similar types of products may have a high cosine similarity even if their total spending differs significantly.
Clustering enables organizations to identify meaningful customer segments and make better business decisions.
Businesses divide customers into groups based on purchasing behavior, demographics, income, interests, or spending patterns.
Marketing teams design campaigns for specific customer groups instead of sending identical promotions to everyone. E-commerce companies recommend products by identifying customers with similar purchasing behavior. Financial institutions identify unusual transactions that do not belong to normal customer clusters. Hospitals group patients with similar symptoms to improve diagnosis and treatment planning. Organizations discover customer preferences and identify new market opportunities. A national retail company wants to improve its marketing strategy. The company collects customer information including: Instead of manually categorizing customers, the company applies a clustering algorithm. The algorithm automatically identifies several customer groups, including: The marketing department creates personalized campaigns for each segment, increasing customer engagement, improving conversion rates, and maximizing return on marketing investment. Continue to Part 2, where you will learn the most popular clustering algorithms, including K-Means Clustering, Hierarchical Clustering, and DBSCAN, along with the Elbow Method, Silhouette Score, Python implementation using Scikit-learn, and customer segmentation examples. Different clustering algorithms use different techniques to group similar data points. The choice of algorithm depends on the type of dataset, the business objective, the number of features, and the shape of the clusters. The three most widely used clustering algorithms in Machine Learning are K-Means Clustering, Hierarchical Clustering, and DBSCAN. K-Means Clustering is the most popular clustering algorithm used for customer segmentation. It divides data into K clusters, where K is chosen by the analyst before training the model. The algorithm assigns each observation to the nearest cluster center (called the centroid) and repeatedly updates the centroids until the clusters become stable. K-Means is simple, fast, and works well for large datasets with clearly separated clusters. Example: A supermarket divides customers into four groups based on annual income and spending score to create personalized promotional campaigns. Hierarchical Clustering creates a hierarchy of clusters instead of assigning all observations directly into a fixed number of groups. The results are usually visualized using a Dendrogram, which shows how clusters are merged or divided. There are two approaches: Unlike K-Means, Hierarchical Clustering does not require specifying the number of clusters in advance. Business Example: A hospital groups patients based on medical history and symptoms to identify disease patterns. DBSCAN stands for Density-Based Spatial Clustering of Applications with Noise. Instead of using centroids, DBSCAN groups observations based on data density. Areas containing many nearby observations form clusters, while isolated observations are treated as noise or outliers. DBSCAN is particularly useful when clusters have irregular shapes. Advantages of DBSCAN: Business Example: A bank detects unusual credit card transactions that do not belong to any normal customer cluster. One of the biggest challenges in clustering is determining the correct number of clusters. Choosing too few clusters may combine different customer groups, while choosing too many clusters may create unnecessary complexity. Two commonly used evaluation techniques are the Elbow Method and the Silhouette Score. The Elbow Method helps determine the optimal number of clusters by measuring the Within-Cluster Sum of Squares (WCSS). The algorithm is trained using different values of K. As K increases, WCSS decreases. The best value of K is usually found where the curve forms an “elbow.” Beyond this point, adding more clusters provides only small improvements. The Silhouette Score measures how well each observation fits within its assigned cluster. Analysts often use the Elbow Method together with the Silhouette Score to select the most appropriate number of clusters. Python provides the Scikit-learn library for implementing clustering algorithms. This program groups customers into three different clusters based on their age and annual income. The scatter plot visually displays customers belonging to different clusters using different colors. An online fashion retailer wants to improve its marketing strategy. The company collects customer information including: Using K-Means Clustering, the retailer identifies four customer groups: Each group receives different promotional offers, resulting in higher customer engagement and improved conversion rates. Business Problem A shopping mall wants to understand customer purchasing behavior. Dataset Features Objective Identify customer groups for targeted marketing campaigns. Expected Outcome This project demonstrates how clustering transforms raw customer data into meaningful business insights. Continue to Part 3, where you will learn the advantages and limitations of clustering, best practices, real-world case studies, frequently asked questions, lesson summary, and key takeaways for customer segmentation. Clustering is one of the most powerful exploratory data analysis techniques in Machine Learning. It helps organizations discover hidden customer groups, identify behavioral patterns, and improve decision-making. However, like every Machine Learning algorithm, clustering has both strengths and limitations. A large online retail company wants to improve customer engagement and increase sales. The company collects customer information from its e-commerce platform, including: The Data Analytics team applies the K-Means clustering algorithm and identifies five distinct customer segments. After implementing targeted marketing campaigns for each segment, the company experiences higher customer engagement, improved conversion rates, increased customer retention, and more efficient marketing spending. Clustering is one of the most valuable techniques in Unsupervised Machine Learning because it automatically identifies meaningful groups within data without requiring predefined labels. Businesses use clustering to segment customers, understand purchasing behavior, detect fraud, improve recommendation systems, and support strategic decision-making. Algorithms such as K-Means, Hierarchical Clustering, and DBSCAN each offer unique strengths depending on the dataset and business objective. By combining appropriate distance measures, cluster evaluation techniques, and domain knowledge, organizations can transform raw data into actionable business insights that improve customer satisfaction and operational efficiency. Clustering is an Unsupervised Machine Learning technique that groups similar observations together based on shared characteristics without using labeled data. Because the algorithm learns patterns from data without predefined target labels or correct answers. K-Means is the most widely used clustering algorithm because it is simple, fast, and performs well on many business datasets. Customer segmentation is the process of dividing customers into groups with similar characteristics so businesses can create personalized marketing strategies and improve customer relationships. Businesses use clustering for customer segmentation, targeted marketing, fraud detection, recommendation systems, healthcare analytics, market research, and demand forecasting. Classification predicts predefined categories using labeled data, while clustering automatically discovers hidden groups without labeled training data. In the next lesson, you will learn Introduction to Deep Learning and Neural Networks. You will explore artificial neural networks, neurons, hidden layers, activation functions, forward propagation, backpropagation, and understand how deep learning powers technologies such as image recognition, speech recognition, recommendation systems, and generative AI.Product Recommendations
Fraud Detection
Healthcare Analytics
Market Research
Real-World Example
Key Takeaways
Popular Clustering Algorithms
K-Means Clustering
How K-Means Works
Hierarchical Clustering
DBSCAN (Density-Based Spatial Clustering)
Choosing the Right Clustering Algorithm
Algorithm
Best For
Main Advantage
K-Means
Large numerical datasets
Fast and simple
Hierarchical
Small and medium datasets
Easy visualization using dendrograms
DBSCAN
Irregular clusters and outlier detection
No need to define K
Choosing the Optimal Number of Clusters
Elbow Method
Silhouette Score
Python Implementation Using Scikit-learn
K-Means Clustering Example
import pandas as pd
from sklearn.cluster import KMeans
data = {
'Age':[22,25,30,35,40,45,50,28],
'Income':[30000,40000,50000,70000,80000,90000,100000,45000]
}
df = pd.DataFrame(data)
model = KMeans(
n_clusters=3,
random_state=42
)
df['Cluster'] = model.fit_predict(
df[['Age','Income']]
)
print(df)
Visualizing Customer Clusters
import matplotlib.pyplot as plt
plt.scatter(
df['Age'],
df['Income'],
c=df['Cluster'],
cmap='viridis'
)
plt.xlabel("Age")
plt.ylabel("Income")
plt.title("Customer Segments")
plt.show()
Business Example: Customer Segmentation
Cluster
Customer Type
Cluster 1
Premium Customers
Cluster 2
Budget Shoppers
Cluster 3
Frequent Buyers
Cluster 4
Occasional Customers
Mini Project: Customer Segmentation Using K-Means
Advantages, Limitations and Best Practices
Advantages of Clustering
Limitations of Clustering
Best Practices for Clustering
Real-World Case Study
Customer Segment
Business Strategy
Premium Customers
VIP rewards and exclusive offers
Frequent Buyers
Loyalty programs and early product access
Budget Shoppers
Discount campaigns and seasonal promotions
Occasional Customers
Personalized reminder emails and coupons
At-Risk Customers
Retention campaigns and special incentives
Key Takeaways
Lesson Summary
Frequently Asked Questions (FAQs)
What is clustering in Machine Learning?
Why is clustering called an unsupervised learning technique?
Which clustering algorithm is most commonly used?
What is customer segmentation?
How do businesses use clustering?
What is the difference between clustering and classification?
What’s Next?