Page 190 - AI Ver 3.0 class 10_Flipbook
P. 190
As a result, the app clusters the photos into categories like "vacation," "family gatherings," or "pets," without
needing any input or labels from the user.
This example shows how unsupervised learning works by discovering hidden patterns in unlabelled data, just like
the child in the swimming pool explores and learns independently.
Let us take an example of the Unsupervised Learning - Fraud Detection:
A bank processes a large number of transactions daily, maintaining a database with details like:
• Transaction amount
• Transaction location
• Time of transaction
• Account activity patterns
The goal is to identify potentially fraudulent transactions. However, the transactions are not pre-labelled as
"fraudulent" or "non-fraudulent."
An Unsupervised learning algorithm is applied to analyse the data and group transactions based on patterns.
The algorithm automatically identifies unusual behaviours, such as:
• Unusually large transactions.
• Purchases from unusual locations.
• Multiple transactions within a short time frame.
The model forms two main clusters:
• Cluster 1: Regular transactions that follow typical patterns.
• Cluster 2: Suspicious transactions that deviate significantly from normal behaviour.
Based on these clusters, the bank can flag transactions in Cluster 2 for further investigation, without needing prior
labels for fraudulent activity.
Regular Transactions
Total No. Total No. of 80 Suspicious Transactions
Transaction
of Regular Suspicious
ID 60
Transactions Transactions Unsupervised
Learning Transaction Frequency 40
Model 20
0
50 100 150 200 250
Transaction Amount ($)
This example illustrates how unsupervised learning helps uncover hidden patterns in unlabelled data, enabling
more effective decision-making.
21 st Century #Media Literacy
Skills
Video Session
Watch the video on "Neural Networks" at the given link:
https://www.youtube.com/watch?v=jmmW0F0biz0 or scan the QR code and answer
the following question:
What are the five things that you should know about neural networks?
188 Touchpad Artificial Intelligence (Ver. 3.0)-X

