Chapter 4

Unlike Supervised Learning, here we don’t categorize

Machine Learning Models label the input!

Unsupervised Learning
Group 1

Group 2

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Unsupervised Learning Models

This time we need to ask machine to label our
data
Features

Sepal Sepal Pedal Pedal

Length Width Length Width
Setosa
Versicolor Samples
Virginica

Labels

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Unsupervised Learning Models

But How?!
Color

Size Similarity

Season

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Unsupervised Learning Models

Similarity
Between all Unlabeled Samples

Distance

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Unsupervised Learning Models

Clustering Techniques

Clustering

Centroid Based Hierarchical Density Based

K-Mean

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean Partitioning the Samples into

K clusters

Step 1 Defining Number of Clusters (k)

k=2

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean

Step 2 Randomly Centroid Selection

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean

Step 3 Assign Each Point to the Nearest Centroid

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean

Step 4 Update the centroid

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean

Step 5 Repeating Previous Step until the Centroid does not Change Anymore

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean
How to Calculate Distance?

 Euclidean
(A line connecting two points)

 Manhattan (Taxicab Geometry)

Using horizontal and vertical lines

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean
Evaluation Metric

𝑛𝑛 𝑚𝑚
2 2
� d 𝑖𝑖, c1 + � d 𝑗𝑗, c2 d(i, c1 )
𝑖𝑖=1 𝑗𝑗=1 d(j, c2 )

Inertia

Smaller inertia

Better Clustering

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

K-Mean
Evaluation Metric
Elbow Method inertia

Elbow Point ≈ Optimal Point

Elbow Point

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Unsupervised Learning Models

Clustering Techniques

Clustering

Centroid Based Hierarchical Density Based

K-Mean DBSCAN

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Unsupervised Learning Models

DBSCAN Density-Based Spatial Clustering of Applications with Noise
Core
Noise
Parameters: eps

• eps

• MinPoints = 4
If No. of Points ≥ MinPoints Core

If No. of Points ≤ MinPoints Border

Border

There are no other points Noise

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Unsupervised Learning Models

Clustering Techniques

Clustering

Centroid Based Hierarchical Density Based

K-Mean Linkage DBSCAN

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7
Unsupervised Learning Models
Hierarchical Clustering Dendrogram

Bottom-up approach

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi
 Chapter 7

Hierarchical Clustering Complete Linkage

Linkage Method

Average Linkage

Single Linkage

Data Science & Machine Learning A-Z: Hands on Python Instructor: Navid Shirzadi