Page 124 - Ai_C10_Flipbook
P. 124
Features of the data:
• Weight: The mass of the fruit.
• Size: The shape or dimension (e.g., medium, long, round, small).
The model is trained with labelled data where the weights and sizes are associated with specific fruit types.
The model learns to recognise patterns and relationships between the features (weight and size) and the label
(fruit type). So, if the trained model is given new data (e.g., Weight: 120 grams, Size: Long), it predicts the fruit
type (Banana) based on the training it received. This example demonstrates how labelled data (weight, size, and
corresponding fruit type) helps the model learn and make predictions for new instances.
Labelled Data
Prediction Output
AI Model Apple
Apple Banana
Grapes
Orange Grapes Banana
Orange
Test Data
Unsupervised Learning
Unsupervised learning approach works on unlabelled dataset. This means that the data which is fed to the machine
is random and there is no knowhow available about it to the model. The machine analyses the data and identifies
patterns, structures, or relationships on its own without any guidance. The goal is to group or organise data based
on similarities or differences.
In this model the major features are identified by the machine, which help the user in understanding the data.
For example, in the data of 100 cat images, if you want to understand some pattern in the data, you would need
to feed this data into the unsupervised learning model and train the machine. Once trained, the machine would
identify patterns in the data. These patterns might already be known to the user, like colour or size, or different
features of the cats.
Unsupervised learning helps discover hidden patterns in unlabelled data. For example, imagine a photo gallery
app that automatically organises a user's photos based on their content. The photos aren’t pre-labelled as "family,"
"vacation," "pets," or "friends." Instead, the app uses unsupervised learning to analyse the photos and group them
based on similarities, such as recognising the same faces, landscapes, or objects like pets or vehicles. 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.
122 Artificial Intelligence Play (Ver 1.0)-X

