Regression Analysis

Regression Analysis is a statistical and machine learning method used to study and model the
relationship between a dependent variable and one or more independent variables. It helps
in predicting continuous numerical values such as price, temperature, sales, or salary.
In regression analysis, we try to understand how changes in independent variables affect the
dependent variable. The goal is to build a mathematical model that can estimate or predict the
value of the dependent variable based on the given inputs.
Definition
Regression analysis is a technique used to estimate the relationship between variables and
predict future outcomes using historical data.
It is widely used in statistics, economics, finance, business analytics, and machine learning.
Key Components of Regression Analysis
• Dependent Variable (Y)
The dependent variable is the target or output variable that we want to predict.
Example:
• House price
• Student marks
• Sales revenue
• Independent Variables (X)
Independent variables are the input variables or predictors that influence the dependent
variable.
Example:
• House size
• Number of bedrooms
• Advertising budget
• Regression Model
The regression model is the mathematical equation that represents the relationship
between X and Y.
• Regression Line (Best Fit Line)
In linear regression, the relationship between variables is represented by a straight line
that best fits the data points.
Example: Predicting House Price
Suppose we want to predict house price based on house size.
House Size (sq ft) Price ($)
1000 150000
1500 200000
2000 250000

Here:
• Independent Variable (X) → House Size
• Dependent Variable (Y) → Price
From this data, regression analysis tries to find a mathematical relationship between size and
price.
For example, the model might learn the equation:
Price=50000+100×SizePrice = 50000 + 100 \times SizePrice=50000+100×Size

Using this equation, if a house size is 1600 sq ft, the predicted price would be:
Price=50000+(100×1600)=210000Price = 50000 + (100 \times 1600) =
210000Price=50000+(100×1600)=210000

Thus, regression helps in estimating the house price for sizes not present in the dataset.
Purpose of Regression Analysis
Regression analysis is used for several purposes:
• Prediction
Estimate future values based on historical data.
• Understanding Relationships
Identify how different variables influence the target variable.
• Trend Analysis
Analyze patterns and trends in data.
• Decision Making
Help businesses and organizations make informed decisions.
Applications of Regression Analysis
Regression analysis is widely used in many real-world fields:
• Business – Sales forecasting and demand prediction
• Finance – Stock price prediction and risk analysis
• Healthcare – Predicting disease risk based on patient data
• Real Estate – Predicting house prices
• Marketing – Measuring the impact of advertising on sales
Advantages of Regression Analysis
• Simple and easy to interpret
 • Helps in understanding relationships between variables
• Useful for prediction and forecasting
• Widely applicable in many fields
Limitations of Regression Analysis
• Works best when relationships are linear
• Sensitive to outliers in data
• Requires sufficient and accurate data

2. Regression
Regression is a machine learning technique used to predict continuous numerical values
based on one or more input variables. Unlike classification, where the output is categorical,
regression predicts a quantitative value.
It works by estimating the relationship between independent variables (inputs) and the
dependent variable (output). Once the relationship is learned from historical data, it can be
used to predict values for new or unseen inputs.
Key Points About Regression
• Dependent Variable (Y) – the output we want to predict.
Example: Salary, house price, temperature, sales.
• Independent Variable(s) (X) – input variable(s) used to make predictions.
Example: Years of experience, house size, advertising budget.
• Regression Model – a mathematical equation that represents the relationship between
inputs and output.
• Prediction – after training, the model can predict output values for new input data.
Example: Predicting Salary Based on Experience
Suppose we have the following data:
Experience (Years) Salary ($)
1 30,000
3 45,000
5 60,000

Objective: Predict the salary of a new employee with 4 years of experience.

Step 1: Identify Variables
• Independent variable → Experience (Years)
• Dependent variable → Salary ($)
Step 2: Find the Relationship
We can assume a linear relationship between experience and salary. A simple linear regression
equation:
Salary=a+b×Experience

Where:
• a = intercept (base salary when experience = 0)
• b = slope (increase in salary per year of experience)

Step 3: Estimate Parameters

From the data, the model might learn:
Salary=15000+9000×Experience

Step 4: Make Predictions

For an employee with 4 years of experience:
Salary=15000+9000×4=51,000

So the predicted salary is $51,000.

Applications of Regression
• House Price Prediction – predicting prices based on size, location, and amenities.
• Stock Price Prediction – predicting stock values over time.
• Sales Forecasting – predicting future sales based on marketing spend, trends, or
seasonality.
• Weather Prediction – estimating temperature, rainfall, or humidity based on historical
patterns.
• Healthcare – predicting patient outcomes, blood sugar levels, or recovery times.
Advantages of Regression
• Simple and easy to interpret.
• Helps predict numerical outcomes.
• Useful for trend analysis and forecasting.
• Can handle multiple independent variables in multiple regression.
Limitations of Regression
• Assumes a specific relationship (e.g., linear for simple linear regression).
 • Sensitive to outliers.
• Requires sufficient and accurate data for reliable predictions.

3. Linear Regression
Linear Regression is the simplest form of regression and is used to model the linear
relationship between an independent variable (X) and a dependent variable (Y).
It assumes that the change in the dependent variable is proportional to the change in the
independent variable, meaning the relationship can be represented as a straight line on a graph.
Definition
Linear Regression is a statistical method that predicts the value of a dependent variable based
on one or more independent variables using a linear equation.
It is widely used in machine learning, data analysis, finance, and business forecasting.
Linear Regression Equation
The general equation of a linear regression line is:
Y=a+bX

Where:
• Y = Predicted value (dependent variable)
• X = Input variable (independent variable)
• a = Intercept (value of Y when X = 0)
• b = Slope (rate at which Y changes for a one-unit change in X)
Interpretation:
• Intercept (a): Base value of Y when X = 0
• Slope (b): How much Y increases (or decreases) for each unit increase in X
The goal of linear regression is to find the best-fitting straight line that minimizes the
difference between predicted and actual values (typically using the least squares method).
Example: Predicting House Price Using Size
Suppose we have the following relationship between house size (X) and house price (Y):
Regression Equation:
Price=50,000+100×Size

Where:
• 50,000 = base price of a house (intercept)
• 100 = price increase per square foot (slope)
Step 1: Predict Price for a Specific Size
If the house size = 1500 sq ft, then:
Price=50,000+(100×1500) =200,000

So the predicted house price is $200,000.

Step 2: How the Model Works
• Collect Data: Historical data of house sizes and prices.
• Compute Regression Parameters: Calculate intercept (a) and slope (b).
• Fit the Line: Draw the best-fitting straight line through the data points.
• Make Predictions: Use the equation to estimate house prices for new sizes.
Advantages of Linear Regression
• Simple and easy to interpret.
• Helps understand the relationship between variables.
• Useful for prediction and forecasting.
• Works well when the relationship between variables is approximately linear.
Limitations of Linear Regression
• Assumes a linear relationship; cannot capture curves or complex patterns.
• Sensitive to outliers that can distort the line.
• Cannot handle multicollinearity (high correlation between independent variables) in
multiple regression.
Applications of Linear Regression
• Predicting house prices based on size, location, and features.
• Forecasting sales revenue based on advertising spend.
• Estimating student marks based on study hours.
• Predicting temperature trends over time.

Visual Representation
• The X-axis represents the independent variable (Size)
• The Y-axis represents the dependent variable (Price)
• The regression line shows the predicted values
All points closer to the line indicate better predictions.
4. Simple Linear Regression
Simple Linear Regression uses one independent variable to predict a dependent variable.
Equation
Y=a+bX
Example
Predicting student marks based on study hours.
Study Hours Marks
2 40
4 55
6 70
The model learns a straight line relationship between study hours and marks.
5. Multiple Linear Regression
Multiple Linear Regression (MLR) is an extension of simple linear regression where two or
more independent variables are used to predict a single dependent variable.
Unlike simple linear regression, which only considers one factor, multiple linear regression
considers multiple factors simultaneously to provide a more accurate prediction.
Definition
Multiple Linear Regression is a statistical and machine learning technique used to model the
linear relationship between several independent variables and a dependent variable.
The model assumes that the dependent variable changes linearly with respect to each
independent variable.
Equation
The general equation for multiple linear regression is:
Y=a+b1X1+b2X2+⋯+bnXn

Where:
• Y = Dependent variable (output to predict)
• X1, X2, … Xn = Independent variables (predictors)
• a = Intercept (value of Y when all X = 0)
• b1, b2, … bn = Coefficients/slopes (amount Y changes per unit change in each X)
Example: Predicting House Price
Suppose we want to predict house price based on three features:
Size (sq ft) Bedrooms Age (Years) Price ($)
1200 2 10 180,000
1500 3 8 220,000
2000 4 5 300,000

Step 1: Identify Variables

• Independent Variables (X):
• Size
• Bedrooms
• Age
• Dependent Variable (Y):
• Price
Step 2: Model Equation
The regression equation might look like:
Price=50,000+100×Size+10,000×Bedrooms−2,000×AgePrice = 50,000 + 100 \times Size + 10,000 \times
Bedrooms - 2,000 \times AgePrice=50,000+100×Size+10,000×Bedrooms−2,000×Age

Interpretation:
• Intercept (50,000): Base price of a house
• Size (100 × Size): For every 1 sq ft increase, price increases by $100
• Bedrooms (10,000 × Bedrooms): Each extra bedroom adds $10,000
• Age (-2,000 × Age): Each year of age reduces the price by $2,000
Step 3: Predict Price
Predict the price of a house with:
• Size = 1500 sq ft
• Bedrooms = 3
• Age = 8 years
Price=50,000+(100×1500)+(10,000×3)−(2,000×8)=214,000

So, the predicted house price is $214,000.

Advantages of Multiple Linear Regression

• Considers multiple factors for more accurate predictions.
 • Helps identify the impact of each independent variable on the dependent variable.
• Useful in business forecasting, real estate, finance, and healthcare.

Limitations of Multiple Linear Regression

• Assumes a linear relationship between variables.
• Sensitive to outliers.
• Requires careful checking for multicollinearity (when independent variables are highly
correlated).
• Complex with many variables, making interpretation harder.

Applications
• Real Estate: Predict house prices based on size, location, age, and amenities.
• Finance: Predict stock prices based on multiple market indicators.
• Marketing: Predict sales based on advertising, pricing, and promotions.
• Healthcare: Predict patient recovery times based on age, medication, and medical
history.
6. Polynomial Regression
Polynomial Regression is an extension of linear regression that is used to model nonlinear
relationships between the independent variable(s) and the dependent variable.
Unlike linear regression, which fits a straight line to the data, polynomial regression fits a
curved line that can capture more complex patterns in the data.
It is especially useful when the relationship between variables cannot be approximated well by
a straight line.
Definition
Polynomial Regression is a type of regression in which the dependent variable (Y) is expressed
as a polynomial function of the independent variable (X).
It allows the model to capture curves, acceleration, or deceleration trends in the data.
Equation
The general form of a polynomial regression equation is:
Y=a+bX+cX2+dX3+⋯+nXk

Where:
• Y = Dependent variable (output)
 • X = Independent variable (input)
• a, b, c, d … n = Coefficients of the polynomial
• k = Degree of the polynomial (e.g., 2 for quadratic, 3 for cubic)
Interpretation:
• a (intercept): Value of Y when X = 0
• b, c, d … n: Influence of X, X², X³, etc., on Y
• Higher-degree terms allow the model to capture curves in the data.
Example: Sales vs. Advertising Budget
Suppose a company tracks how sales increase with advertising budget:
Advertising Budget (X) Sales (Y)
1 10
2 25
3 60
4 120

Observations:
• Sales do not increase linearly with advertising.
• The increase accelerates as the budget grows.
Step 1: Fit a Polynomial Model
We can fit a quadratic (degree 2) or cubic (degree 3) polynomial to capture the curvature:
Sales=a+b×Budget+c×Budget2

After training, the model might find:

Sales=2+3×Budget+5×Budget2

Step 2: Make Predictions

For a new advertising budget of X = 3.5:
Sales=2+3×3.5+5×(3.5)2=73.75

Predicted sales = 73.75 units.

Step 3: Visualization
• X-axis: Advertising budget
• Y-axis: Sales
• The polynomial regression curve fits the data points better than a straight line, showing
 the nonlinear trend.
Advantages of Polynomial Regression
• Can model nonlinear relationships.
• Captures curved trends in data that linear regression cannot.
• More flexible for complex patterns.
Limitations of Polynomial Regression
• Higher-degree polynomials can overfit the data.
• Sensitive to outliers, which can distort the curve.
• Difficult to interpret as degree increases.
• Requires careful selection of the polynomial degree.
Applications
• Sales forecasting: Modeling nonlinear growth of sales with advertising.
• Economics: Modeling price-demand curves.
• Engineering: Modeling physical phenomena with curves (e.g., stress-strain
relationships).
• Environmental Science: Modeling temperature or pollution trends over time.
7. Sparse Model
A Sparse Model is a type of machine learning model where most of the feature coefficients are
zero, meaning only a small subset of features contributes significantly to the predictions.
In other words, the model ignores irrelevant or unimportant features, which makes it simpler,
faster, and more interpretable. Sparse models are particularly useful when dealing with high-
dimensional datasets (datasets with many features).
Definition
A sparse model is one in which most predictors have zero or negligible weights, leaving only
the important features to influence the outcome. This reduces complexity and makes the model
easier to understand.
• Non-zero coefficients: Features that matter for prediction
• Zero coefficients: Features that are ignored
Why Sparse Models Are Useful
• Reduce Model Complexity
By ignoring irrelevant features, the model becomes simpler and faster to compute.
• Improve Interpretability
With fewer features contributing to predictions, it is easier to understand why the model
makes certain predictions.
 • Prevent Overfitting
Using only important features reduces the risk of overfitting to noise in the data.
• Efficient Storage and Computation
Sparse models require less memory and are faster to evaluate.
How Sparse Models Are Created
Sparse models are often created using regularization techniques that penalize unnecessary
complexity:
• Lasso Regression (L1 Regularization)
• Adds a penalty proportional to the absolute value of coefficients.
• Encourages many coefficients to become exactly zero.
• Feature Selection Methods
• Automatic methods can remove irrelevant features before training.
Example: Predicting Disease Using 100 Medical Features
Suppose we have a dataset with 100 features such as age, blood pressure, cholesterol, genetic
markers, etc., to predict the risk of a disease.
After training a sparse model using Lasso Regression:
• Only 5 features are found to be significant.
• The other 95 features have coefficients equal to zero.
This means the model focuses only on the 5 important medical features for prediction,
ignoring the irrelevant ones.
Step 1: Original Dataset
Patient Feature1 Feature2 … Feature100 Disease Risk
P1 45 120 … 0.7 High
P2 50 130 … 0.5 Low
… … … …… …

Step 2: After Sparse Model Training

• Only Feature2, Feature7, Feature12, Feature33, Feature55 have non-zero coefficients.
• All other features have zero coefficients and are ignored.
Result: The model becomes simpler and focuses only on the most important predictors.
Advantages of Sparse Models
• Reduces overfitting by ignoring noise features.
• Makes interpretation easier for humans.
• Improves computational efficiency for high-dimensional data.
 • Useful in medical diagnosis, text classification, genetics, and finance.
Applications
• Healthcare: Identifying key biomarkers from hundreds of possible medical features.
• Text Mining: Selecting important words/features for document classification.
• Finance: Using only critical financial indicators for risk prediction.
• Genomics: Identifying genes most relevant for a particular disease.

8. Unsupervised Learning
Unsupervised Learning is a type of machine learning where the model learns from data that
has no labeled outputs. Unlike supervised learning, where each data point has a known target,
in unsupervised learning the algorithm must discover patterns, structures, or relationships in
the data on its own.
The main goal of unsupervised learning is to identify hidden structures in the data, such as
clusters, associations, or anomalies.
Definition
Unsupervised learning is a machine learning approach where:
• The data has no labels (no predefined output).
• The algorithm explores the data to identify patterns, groupings, or relationships.
• It is used for data exploration, pattern recognition, and dimensionality reduction.

Common Applications of Unsupervised Learning

• Customer Segmentation
Group customers with similar behavior for targeted marketing.
• Market Analysis
Identify patterns in consumer behavior or sales trends.
• Image Compression
Reduce image data by finding patterns and representing them efficiently.
• Anomaly Detection
Detect unusual data points, such as fraud detection in banking.
• Recommendation Systems
Suggest items based on similarities between users or items.
Key Concepts
• Patterns and Structure Discovery
The algorithm looks for similarities and differences in the dataset.
• Clustering
 Grouping similar data points together.
• Dimensionality Reduction
Reducing the number of features while retaining important information (e.g., PCA).
• Distance and Similarity Measures
Determine how “close” or “similar” data points are.
Example: Customer Segmentation
Suppose a company has the following customer data:
Age Spending ($)
25 200
40 500
22 180
45 520
Objective: Group customers with similar spending behavior.
Step 1: Choose an Algorithm
Common algorithms for clustering in unsupervised learning:
• K-Means – assigns customers to k clusters based on similarity.
• Hierarchical Clustering – builds a tree of clusters.
• DBSCAN – detects clusters of varying shapes and identifies outliers.
Step 2: Apply Clustering
Suppose we use K-Means with k = 2 clusters:
• Cluster 1 → Low Spenders:
Customers aged 22 and 25 → Spending 180–200
• Cluster 2 → High Spenders:
Customers aged 40 and 45 → Spending 500–520
The algorithm automatically discovers these clusters without any labels.
Step 3: Use the Results
• Targeted Marketing – Offer discounts to low spenders and premium offers to high
spenders.
• Business Strategy – Adjust product offerings based on customer groups
Advantages of Unsupervised Learning
• Can discover hidden patterns in unlabeled data.
• Helps in data exploration and feature understanding.
• Useful in market analysis, anomaly detection, and recommendation systems.
Limitations
 • Difficult to evaluate performance, as there are no labels.
• Sensitive to parameter choices (e.g., number of clusters in K-Means).
• May detect patterns that are not meaningful without domain knowledge.

9. Clustering
Clustering is a core technique in unsupervised learning where the goal is to group similar
data points together. Each group of similar points is called a cluster.
Unlike supervised learning, no labeled data is provided; the algorithm identifies natural
patterns or groupings in the data automatically.
Definition
Clustering is an unsupervised learning method that divides a dataset into distinct groups
based on similarity or distance between data points.
• Cluster: A collection of data points that are similar to each other.
• Goal:
• Points in the same cluster → highly similar
• Points in different clusters → dissimilar
Why Clustering Is Important
• Discover hidden patterns in unlabeled data.
• Group data for segmentation, pattern recognition, and data summarization.
• Often used as a preprocessing step for other machine learning tasks.
Common Clustering Algorithms
• K-Means Clustering
• Divides data into k clusters based on distance to the cluster centroid.
• Minimizes within-cluster variance.
• Simple and efficient for large datasets.
• Hierarchical Clustering
• Builds a tree-like structure (dendrogram) showing nested clusters.
• Can be agglomerative (bottom-up) or divisive (top-down).
• DBSCAN (Density-Based Spatial Clustering of Applications with Noise)
• Groups points based on density.
• Can detect clusters of arbitrary shape and identify outliers.
Example: Customer Segmentation
Suppose we have the following customer data:
Customer Annual Spending ($)
A 200
B 220
C 900
D 950

Step 1: Apply a Clustering Algorithm

Using K-Means with k = 2, the algorithm groups the customers based on spending behavior:
• Cluster 1 → Low spenders:
Customers A (200) and B (220)
• Cluster 2 → High spenders:
Customers C (900) and D (950)
Step 2: Interpretation
• Cluster 1 represents customers who spend less annually → may benefit from discount
offers.
• Cluster 2 represents high-value customers → can be targeted with premium products or
loyalty programs.
The clustering algorithm automatically identifies groups without prior knowledge of the
spending patterns.
Step 3: Visualization
• X-axis: Customer
• Y-axis: Annual spending
• Clusters: Points close together belong to the same cluster; distant points belong to
different clusters.
A plot would show two distinct clusters separated by spending levels.
Advantages of Clustering
• No labeled data required → useful for exploring new datasets.
• Groups similar data for analysis and decision-making.
• Can reduce dataset complexity by summarizing data into clusters.
Limitations of Clustering
• Choosing the number of clusters (k) can be subjective.
• Sensitive to outliers and noise.
• Performance depends on the distance metric used.
 • Clusters may not always have clear boundaries.
Applications
• Marketing: Customer segmentation based on spending, preferences, or behavior.
• Healthcare: Group patients based on symptoms or genetic markers.
• Finance: Detect fraudulent transactions or group investments.
• Image Processing: Group pixels in image segmentation.

10. Similarity and Distance

In clustering and unsupervised learning, one of the most important tasks is to measure how
similar or different data points are. This is done using distance or similarity metrics.
The idea is simple:
• Points that are close together → highly similar → likely to be in the same cluster
• Points that are far apart → dissimilar → likely to belong to different clusters
Choosing the right distance metric is critical for the performance of clustering algorithms like
K-Means, Hierarchical Clustering, and DBSCAN.
Common Distance Measures
1. Euclidean Distance
Euclidean distance is the most commonly used metric. It measures the straight-line distance
between two points in a multi-dimensional space.
Formula (2D case):
d=sqrt((x1−x2)2+(y1−y2)2)
Where:
• (x1,y1) and(x2,y2) are the coordinates of the two points.
Example:
• Point A = (2, 3)
• Point B = (5, 7)
d=sqrt((5−2)2+(7−3)2)=5
Interpretation:
The Euclidean distance between A and B is 5 units. In clustering, a smaller distance means the
points are more similar.
Use Cases:
• K-Means clustering
• Image segmentation
 • General continuous data similarity
2. Manhattan Distance
Manhattan distance, also called city-block distance or L1 norm, measures the distance along
axes at right angles, like navigating a city grid.
Formula (2D case):
d=∣x1−x2∣+∣y1−y2∣
Where:
• Absolute differences are summed along each dimension.
Example:
• Point A = (2, 3)
• Point B = (5, 7)
d=∣5−2∣+∣7−3∣=7
Interpretation:
The Manhattan distance between A and B is 7 units. It represents the distance if you can only
move along axes, not diagonally.
Use Cases:
• Grid-like or lattice structures
• Certain high-dimensional datasets
• Some clustering algorithms (like DBSCAN with Manhattan distance)
Role in Clustering
• Distance metrics determine how clusters are formed.
• In K-Means, points are assigned to the nearest cluster centroid based on distance.
• In Hierarchical Clustering, distance between points or clusters guides merging or
splitting.
• Choosing the right metric affects cluster shape, size, and accuracy.

11. Quality Measures of Clustering

Clustering quality measures are used to evaluate how well a clustering algorithm groups similar
data points together and separates dissimilar data points into different clusters. Since clustering is
an unsupervised learning technique, there are no predefined labels. Therefore, these measures
help determine whether the clusters formed are meaningful and well-structured.

1. Cohesion (Intra-Cluster Similarity)

Cohesion measures how closely related the data points are within the same cluster. It evaluates
the compactness of a cluster by calculating the distance between the data points belonging to that
cluster.
If the data points inside a cluster are very close to each other, the cluster is considered highly
cohesive.

Key Points

Measures within-cluster similarity.

Lower distance between points indicates better clustering.

High cohesion means data points in a cluster are very similar.

Mathematical Idea

Cohesion is often measured using the sum of squared distances between each point and the
cluster centroid.

Example

Suppose a cluster contains student marks:

[70, 72, 71, 69]

These values are very close to each other. Therefore, the cluster has high cohesion.

If the cluster had values like:

[70, 40, 90, 10]

The distance between points is large, so cohesion is poor.

2. Separation (Inter-Cluster Distance)

Separation measures how distinct or far apart different clusters are from each other. It evaluates
the distance between clusters, usually by measuring the distance between their centroids.

Good clustering should produce clusters that are well separated from each other.

Key Points

Measures distance between clusters.

Larger separation means clusters are clearly distinguishable.

Low separation means clusters may overlap.

Mathematical Idea

Separation is often measured as the distance between cluster centroids.

Example
Consider two clusters:

Cluster A: [10, 12, 11]

Cluster B: [80, 85, 82]

These clusters are far apart, so separation is high.

But if clusters were:

Cluster A: [10, 12, 11]

Cluster B: [13, 15, 14]

They are very close, so separation is low, meaning clustering is poor.

3. Silhouette Score

The Silhouette Score is one of the most commonly used metrics to evaluate clustering quality. It
combines both cohesion and separation into a single value.

It measures how similar a data point is to its own cluster compared to other clusters.

Where:

a = average distance between the point and other points in the same cluster (cohesion)

b = average distance between the point and points in the nearest neighboring cluster (separation)

Range of Silhouette Score

Score Meaning
+1 Data point is perfectly matched with its cluster
0 Data point lies between two clusters
-1 Data point is assigned to the wrong cluster
Interpretation

Score close to +1 → clusters are well separated and compact.

Score around 0 → clusters overlap.

Score close to -1 → incorrect clustering.

Example

Suppose a student belongs to a cluster of students with similar marks.

If their marks are very close to students in their cluster → high silhouette score.

If their marks are similar to another cluster → score near zero.

If they are closer to another cluster → negative score.

12. Case Study Example
Customer Segmentation in Retail
A retail company wants to group customers based on purchasing behavior.
Dataset:
Customer Age Annual Spending
C1 22 200
C2 25 220
C3 40 800
C4 42 850
Using K-Means clustering, the algorithm forms two clusters:
Cluster 1 → Low spenders
(C1, C2)
Cluster 2 → High spenders
(C3, C4)
Business Benefit
The company can:
• Offer discounts to low spenders
• Offer premium products to high spenders