 +91 99908 16780  info@ai-cbse.

com

  

Unit 6: Machine Learning Algorithms

Ai Presentation
Generator

Create beautiful, engaging content

with forma!ing and design work.

aippt.com Open

 

 66

1 1 1

1. Introduction to Machine Learning

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

Machine Learning (ML) is a subset of Arti@cial Intelligence (AI) that focuses on

enabling machines to learn from data and make decisions or predictions without
being explicitly programmed. It contrasts with traditional programming, where
rules are prede@ned. Instead, ML systems develop their own models based on
data inputs and outcomes, allowing them to improve over time.

Real-Life Examples:

Shopping Recommendations: When you browse online stores and

receive suggestions like socks after searching for shoes, that’s ML at
work—analyzing patterns of customer behavior to offer relevant
products.

Entertainment Platforms: Systems like NetRix or Spotify use your

viewing or listening history to recommend similar content, thanks to
ML algorithms analyzing user preferences.

Social Media Facial Recognition: When Facebook suggests tagging a

friend in a photo, ML is recognizing patterns in the image based on
previously tagged photos.

Chatbots: Virtual assistants like Siri or Alexa engage in natural

language conversations, adapting responses based on past
interactions and understanding user queries through ML algorithms.

2. Types of Machine Learning

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

a. Supervised Learning

DeFnition: Supervised learning occurs when a model is trained on a labeled

dataset, meaning that each input has a corresponding output. The model
learns to map inputs to outputs, so it can predict the output for unseen data.

Common Tasks in Supervised Learning:

1. ClassiFcation:

DeFnition: The task of predicting discrete labels (categories) for

new data based on patterns learned from labeled examples.

Examples:

Spam Detection: Classifying emails as “spam” or “not spam”

based on labeled historical email data.

Medical Diagnosis: Predicting whether a patient has a

certain disease (e.g., cancer) based on input features like
age, symptoms, and test results.

Binary ClassiFcation: Problems with two outcomes (e.g.,

Yes/No, True/False).

Multi-Class ClassiFcation: Problems with more than two

categories (e.g., recognizing dog breeds from an image
dataset).

2. Regression:

DeFnition: Regression is used for predicting continuous values. It

is particularly useful in cases where we are predicting quantities
rather than categories.

Examples:

House Price Prediction: Predicting the price of a house

based on features like location, size, and number of
bedrooms.

Stock Price Forecasting: Predicting future stock prices

based on historical price trends.

Key Supervised Learning Algorithms:

Linear Regression: A basic method for regression that @nds a line

(linear relationship) through the data points. It is simple but highly
effective in predicting continuous outcomes like sales forecasts.

Example: A company might use linear regression to predict

future sales based on past marketing expenses.

K-Nearest Neighbors (KNN): A classi@cation algorithm that

predicts the label for a new data point by @nding the ‘k’ most
similar points (neighbors) in the training data.

Example: A healthcare system might use KNN to classify

patients as “high-risk” or “low-risk” for a disease by
comparing them to past patients with similar symptoms.

b. Unsupervised Learning

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

DeFnition: In unsupervised learning, the algorithm works on unlabeled data

(without prede@ned outputs). The goal is to identify hidden patterns,
groupings, or structures within the data.

Common Tasks in Unsupervised Learning:

1. Clustering:

DeFnition: Grouping data points based on similarity. In clustering,

the algorithm identi@es clusters of data points that are more
similar to each other than to points in other clusters.

Examples:

Customer Segmentation: In marketing, companies use

clustering to group customers based on purchasing
behavior, enabling personalized marketing strategies for
each group.

Image Categorization: Grouping similar images together,

such as organizing photos by themes like landscapes,
portraits, or animals.

2. Dimensionality Reduction:

DeFnition: Reducing the number of input variables in a dataset

while retaining important information. This helps simplify the
dataset for analysis and reduces computational costs.

Examples:

PCA (Principal Component Analysis): Used in data science

to reduce a high-dimensional dataset into fewer dimensions
that capture the most variance.

Key Unsupervised Learning Algorithms:

K-Means Clustering: A popular clustering algorithm that groups

data points into ‘k’ clusters based on their similarity. The
algorithm iteratively assigns data points to clusters and adjusts
the cluster centers.

Example: A clothing retailer may use K-Means to group

customers by purchasing patterns, such as budget-
conscious buyers, fashion-conscious buyers, and frequent
shoppers.

c. Reinforcement Learning

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

DeFnition: In reinforcement learning, an agent learns how to act in an

environment by receiving feedback in the form of rewards or penalties. The
agent’s goal is to maximize cumulative rewards through trial and error.

Examples:

Self-Driving Cars: Reinforcement learning helps autonomous vehicles

make decisions (e.g., when to stop, turn, or accelerate) based on the
current environment and maximizing safety.

Game AI: Systems like AlphaGo use reinforcement learning to improve

their gameplay by learning from previous game outcomes and
improving strategies over time.

Key Reinforcement Learning Algorithms:

Q-Learning: A basic reinforcement learning algorithm where the agent

learns to take actions based on maximizing future rewards.

Deep Q-Networks (DQN): Combines Q-learning with deep neural

networks to handle more complex environments.

3. Key Machine Learning Algorithms and Techniques

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

a. Linear Regression (Supervised Learning)

Concept: In linear regression, we predict a continuous outcome based on

the linear relationship between input variables (independent variables) and
the output variable (dependent variable).

Equation:

World Applications:

Predicting Employee Salaries: Based on years of experience and education

level.

Forecasting Sales: A company can predict future sales based on marketing

spend and

b. K-Nearest Neighbors (KNN) (Supervised Learning)

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

Concept: The KNN algorithm classi@es new data points based on the
majority class of their ‘k’ nearest neighbors. It measures similarity using
distance metrics like Euclidean distance.

Steps:
1. Select the number k (number of neighbors).

2. Compute the distance of the new data point to all existing data points.

3. Identify the ‘k’ nearest neighbors.

4. Assign the most common class label to the new data point.

Real-World Applications:

Disease Diagnosis: Classifying patients as “healthy” or “at-risk” based

on symptoms.

Image Recognition: Grouping images of similar objects, such as

grouping pictures of cars, animals, or plants.

c. K-Means Clustering (Unsupervised Learning)

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

Concept: K-Means is a clustering algorithm that partitions data into ‘k’

clusters based on the similarity between data points. Each cluster is
represented by its centroid (center), and data points are assigned to the
cluster with the nearest centroid.

Steps:
1. Choose the number k (clusters).

2. Initialize k random centroids.

3. Assign each data point to the nearest centroid.

4. Recalculate centroids and repeat until convergence.

Real-World Applications:

Market Segmentation: Segmenting customers into distinct groups

based on buying patterns.

Image Segmentation: Partitioning images into distinct segments

based on color or texture.

4. Challenges in Machine Learning

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

While machine learning has become an essential tool for solving complex
problems, it also comes with several challenges that practitioners must navigate
to ensure effective and ethical use. Below are some of the major challenges:

a. OverFtting

DeFnition: Over@tting occurs when a machine learning model becomes too

complex and learns the noise or random Ructuations in the training data,
instead of capturing the underlying patterns.

How It Happens: When a model is too Rexible (e.g., having too many
parameters or overly complex algorithms), it may perform extremely well on
the training data by memorizing speci@c details rather than generalizing the
broader patterns. As a result, it performs poorly on unseen or new data.

Example: Imagine a model designed to predict house prices. If it over@ts, the

model might learn very speci@c details about houses in the training set, like
the exact number of windows or the speci@c address, rather than focusing
on more general factors like size, location, and market trends. While it may
predict perfectly for the training set, it could fail when applied to new data.

Impact: Over@tting leads to poor generalization, where the model performs

well during training but fails to provide accurate predictions in real-world
situations.

Solutions:

Cross-validation: By splitting data into training and validation sets (or

using techniques like k-fold cross-validation), you can test the model’s
performance on unseen data and detect over@tting early.

Simplifying Models: Reducing the model complexity (e.g., using fewer

parameters) can help prevent over@tting.

Regularization: Techniques like L1 or L2 regularization penalize large

coejcients in the model, preventing it from @tting noise in the training
data.

More Training Data: Sometimes, over@tting occurs because the

dataset is too small or imbalanced. Providing more training data helps
the model generalize better.

b. Bias and Fairness

Ai Presentation Generator

Create beautiful, engaging content with formatting and design work.

aippt.com

DeFnition: Machine learning models are highly dependent on the data they
are trained on. If the training data contains biases (e.g., historical biases,
prejudices, or imbalanced representation), the model is likely to inherit and
perpetuate these biases. This raises concerns about fairness, especially in
sensitive domains like hiring, law enforcement, and healthcare.

Types of Bias:
1. Sampling Bias: When the training data does not represent the full
diversity of the population or problem space. For instance, if a facial
recognition system is trained predominantly on images of lighter-
skinned individuals, it may perform poorly on darker-skinned faces.

2. Label Bias: If historical data reRects biased decisions (e.g., biased

hiring practices), then a model trained on that data will replicate those
biases.

3. ConFrmation Bias: This occurs when models are trained on data that
already supports a hypothesis or assumption, reinforcing existing
outcomes without critically evaluating new data.

Example:

Hiring Algorithms: In 2018, it was revealed that an AI hiring tool used

by a major tech company favored male candidates because it had been
trained on historical data from the company, which reRected years of
male-dominated hiring.

Criminal Justice Algorithms: Algorithms used to predict recidivism

(the likelihood of a convicted criminal reoffending) have been criticized
for being biased against minorities due to biased historical arrest and
conviction data.

Impact: Biased algorithms can lead to unfair treatment of individuals or

groups, perpetuating inequalities in areas like hiring, credit approval,
healthcare access, and law enforcement.

Solutions:

Balanced Data: Ensuring that the training data is diverse and

represents the entire population.

Fairness Metrics: Using fairness metrics (e.g., demographic parity,

equalized odds) to assess model fairness during and after training.

Bias Mitigation Techniques: Using techniques like adversarial

debiasing, reweighting data, or modifying model architecture to reduce
biases.

Regular Audits: Conducting regular audits of algorithms and their

outcomes to identify and address biases.

c. Interpretability (“Black Box” Models)

DeFnition: Some machine learning models, particularly complex ones like

deep neural networks, are often referred to as “black boxes” because their
internal workings are dijcult to interpret. While these models may produce
highly accurate predictions, it is challenging to understand how they arrived
at a particular decision.

Why It’s Important: In many @elds, especially those involving critical

decision-making (e.g., healthcare, @nance, or criminal justice), it is essential
to explain why a model made a certain prediction or recommendation. A
lack of interpretability can lead to distrust in the model, particularly when
mistakes are made or when the consequences of a wrong decision are
severe.

Example:

Healthcare: A deep learning model might predict the likelihood of a

patient developing a certain disease with high accuracy. However,
doctors may be hesitant to trust the model’s prediction if they cannot
understand what features (e.g., symptoms, patient history) inRuenced
the decision.

Credit Scoring: In @nancial institutions, models may reject loan

applications without explaining the exact factors leading to the
rejection, making it hard for individuals to challenge or understand the
decision.

Impact: Lack of transparency in AI decision-making can lead to decreased

trust in machine learning systems, particularly in high-stakes applications. It
can also make it dijcult to debug or improve the model.

Solutions:

Model SimpliFcation: Using simpler models (e.g., decision trees, linear

regression) that are easier to interpret, even if they are slightly less
accurate than complex models.

Explainable AI (XAI): Techniques like LIME (Local Interpretable Model-

Agnostic Explanations) or SHAP (SHapley Additive exPlanations)
provide insights into how black-box models make decisions by
approximating local behaviors of the model.

Transparent Algorithms: Employing algorithms like decision trees or

rule-based models that are inherently interpretable and can provide a
clear rationale for their predictions.

Feature Importance Analysis: Analyzing the importance of different

input features in making predictions, which can help explain model
behavior.

Conclusion
Machine learning algorithms are powerful tools that help machines learn from
data, detect patterns, and make predictions in various real-world applications.
Understanding the different types of machine learning—supervised, unsupervised,
and reinforcement learning—allows us to apply the right methods to solve
problems like classi@cation, regression, clustering, and decision-making in
dynamic environments.

ai cbse
This site is dedicated to provide contents, notes,
questions bank,blogs,articles and other materials for AI
students of CBSE.

Related posts

October 1, 2024

UNIT 5 : Data Literacy -Data Collection to Data Analysis

 Read more

September 26, 2024

UNIT 6 : Evaluation

 Read more

September 23, 2024

UNIT 5 : Natural Language Processing

 Read more

Leave a Reply
Your email address will not be published. Required @elds are marked *

Comment *

Name *

Email *

Website

Save my name, email, and website in this browser for the next time I comment.

Post Comment

Search

Enter your search

About our company

At ai-cbse.com, we are dedicated to empowering students and professionals with
knowledge and skills in the @eld of arti@cial intelligence. www.ai-cbse.com As AI
continues to transform our world, we believe that understanding its principles and
applications is essential for navigating the future.

Quick Contact Form

Name: Phone:

E-mail address: Subject:

Message:

The capital of India?

Send message

Latest posts

Model Test Paper XII AI 2024(1)

 January 6, 2025
0

MODEL TEST PAPER (XII AI)

0
 December 29, 2024

Categories

AI Model Test Paper

Articles

Blogs

CBSE Curriculum/Sample Papers

Class 10
Privacy - Terms
Class 11