Advance Concept of Modeling Notes – Class 10 AI (417)

“Want to score well and understand every concept of Advance Concept of Modeling of Class 10 AI (417) easily?”
Don’t worry! These Advanced Concept of Modeling Notes for Class 10 AI will help you build concept clarity and prepare effectively for your CBSE board exams. All key topics are carefully curated as per the latest CBSE syllabus. These AI Modeling Notes of Class 10 are designed in a simple and student-friendly way, making even complex concepts easy to understand. They include important explanations, examples, and exam-oriented points to boost your confidence.

Understand AI, ML and DL

Artificial Intelligence

• Artificial Intelligence (AI) refers to techniques that enable machines to mimic human intelligence.
• AI systems use algorithms and data to analyse information, learn from it, and produce intelligent outputs.

Machine Learning

• Machine Learning (ML), a subset of AI, enables machines to improve their performance with experience.
• By learning from data, ML systems refine their predictions, correct mistakes, and achieve better results over time — without being explicitly programmed.

Examples of Machine Learning

Deep Learning

• Deep Learning (DL), a subset of Machine Learning, uses artificial neural networks inspired by the human brain.
• It automatically learns complex patterns from vast amounts of data and performs specific tasks with high accuracy.

Artificial Neural Network consist of multiple layers such as input layers, hidden layers, and output layers that help ANN automatically learn and recognize complex patterns from data. As the number of layers increases, the network can learn and identify more complex patterns.

Examples of Deep Learning

Difference between AI, ML and DL

Common terminologies used with data

What is data?

• Data is raw facts, figures, text, images given as input to AI/ML model to learn from and make predictions.
• It is information in any form
• Organized in rows and columns

What is Features?

• Features are individual characteristics of data that are used by AI/ML to make decisions
• Columns of the tables are called features
• Some features are special, they are called labels

What is Label?

• Label is the output or target value that AI/ML Model is trained to predict.
• It is the process of assigning meaning to data based on the context of the problem being solved.

Example: Identify Data, Features and Labels from Student Performance Dataset

Labeled Data and Unlabeled Data

Labeled Data – Data to which some tag or label is attached.
Unlabeled Data – Raw data to which no tag is attached.

What is Training data set?

• Training dataset is collection of examples with inputs(features) used to teach the machine learning model.
• The model learns patterns from this data.
• Set of labeled data is used to train model.

What is Testing data set?

• Testing data is a separate set of examples used to test the accuracy of model.
• Testing data is never seen before by model.
• Test is performed without labelled data and then verify results with labels.

MODELING

• AI Modeling refers to developing algorithms (models) to get intelligent outputs.
• It is writing code to make a machine artificially intelligent.

Types of AI Models

 Rule-Based Approach

• AI modeling follows rules defined by developers.
• The machine applies predefined instructions to data to produce output.
• The system does not learn or improve from new data.

Example: Rule-Based Chatbot

• Works on a set of predefined questions and responses
• Uses decision rules (IF condition → THEN action) to reply
• Cannot learn or improve from new situations
• Static models (Once trained, the model cannot improvise)

Limitation of Rule-based approach

• Learning is static (does not change after training)
• Cannot improve or adapt to new data
• Fails when it encounters unseen or different data

Learning-Based Approach

• AI modelling where the machine learns by itself.
• The machine learns from data instead of predefined rules
• It uses input data and feedback (labels/answers) for learning
• The system automatically finds patterns and relationships in data
• It can adapt to new and unseen data
• It is used in Machine Learning and Deep Learning

Example: Spam email filter

• Learns from labelled emails without being explicitly programmed.
• Identifies patterns using features like words, sender, attachments
• Classifies new emails as spam or not spam
• Improves accuracy over time with more data

Difference between Rule based approach and learning based approach

Categories of Machine Learning based models (Learning Based Approach)

Supervised Learning

• Uses labelled dataset (data with correct answers/tags)
• The trainer must know the data to label it correctly
• The model learns by comparing input with correct output
• Similar to a teacher teaching with examples and then testing
• Used to predict or classify new data

Sub categories of Supervised Learning

There are two types of Supervised Learning models:

• Classification model
• Regression model

Classification model

• This model works with discrete data.
• Output is categorical.
• Examples:
  • Email Spam detection
  • Weather prediction
  • Loan approval system
  • Face recognition

Regression model

• This model works with continuous data.
• Output is numerical values
• Examples:
  • Salary prediction
  • Temperature prediction
  • Car price prediction
  • Sales forecasting (predict future sales)

Unsupervised Learning

• Works on unlabelled data (no correct answers/tags given)
• The person training the model may not know the data details
• The machine learns without any supervision or guidance
• Machine analyses data on its own to find patterns, similarities/differences in data.
• Helps in understanding unknown or raw data
• Learning is based on self-exploration
• Similar to a child learning without a teacher

Difference between Supervised and Unsupervised learning

Sub categories of Unsupervised Learning Model

Unsupervised learning models can be further divided into two categories:

• Clustering model
• Association model

Clustering model

• Groups data points based on similar features or patterns
• Data is not labelled in advance (no categories given)
• The model automatically finds hidden patterns in the data
• Examples:
  • Movie recommendation – movies are grouped based on Action, Comedy etc.
  • Customer segmentation – group customers based on buying behaviour.
  • Email spam filtering – emails grouped into spam, important, promotions etc.

Difference between Classification and Clustering

Association model

• Identifies relationships between different variables in a dataset
• Helps understand how one factor influences or is related to another
• Used in product recommendation systems to suggest items based on user behaviour.
• Examples:
  • Online Product Recommendations – Buy a laptop → suggestions for mouse, laptop bag, keyboard
  • Food Ordering Apps – Order a burger → app suggests fries + cold drink
  • Supermarket Offers – Buy shampoo → get conditioner discount

Summary of detailed classification of ML models

Reinforcement Learning

• Machine learns by trying different actions and learning from mistakes (trial and error)
• Receives feedback as reward (right) or penalty (wrong)
• Improves its decisions over time based on past experience
• Does not need labelled data or predefined answers
• Learns by interacting with its environment continuously
• Best suited for real-life situations where conditions keep changing
• Examples:
  • Game Playing – Win games by trying different moves and improving through rewards and penalties
  • Self-Driving Cars – Learns when to brake, turn, or accelerate based on feedback from the environment
  • Robot Navigation – Robot learns the correct path by avoiding obstacles and reaching the goal
  • Online Ads Recommendation – Learns which ads to show to users based on clicks (reward) or no clicks (penalty)
  • Traffic Signal Control – Adjusts signal timing based on traffic flow to reduce congestion

Sub categories of Deep Learning

There are two sub categories of DL:

• Artificial Neural Network (ANN)
• Convolution Neural Network (CNN)

Artificial Neural Network

Artificial Neural networks are modelled on the human brain and nervous system. They are able to automatically extract features without input from the programmer. Every neural network node is essentially a machine learning algorithm. It is useful when solving problems for which the data set is very large.

How ANN works?

• A Neural Network is made up of multiple layers, and each layer contains small units called nodes
• Each node performs a task and passes the result to the next layer
• Neural Network consists of Input layer, Hidden layer and Output layer.

• Input Layer (First Layer):
  • Takes input data and sends it forward to Hidden layer.
  • No processing happens here.
• Hidden Layers (Next to Input Layer):
  • Actual processing happens here. It is not visible to the user.
  • Each node has its own algorithm which it applies on data.
  • Hidden layers process data using weights and biases, and pass it forward after applying an activation function.
• Output Layer(Final Layer):
  • Gives the result/output to the user.
  • The network learns through a trial-and-error process until it produces the correct output.
  • With each try, Weights are adjusted based on the error between predicted and actual output.
• A neural network can have multiple hidden layers depending on problem complexity
• The number of nodes in each layer can vary

Perceptron – How AI Makes Decisions

• A Perceptron is a simple model of a neural network used to make decisions
• It works like human thinking, where multiple factors affect a decision

Inputs (Factors)

• The perceptron takes multiple inputs (X1, X2, X3, …)
• Example factors:
  • Do I have a jacket?
  • Do I have an umbrella?
  • Is it sunny now?
  • What is the weather forecast?
• These inputs are converted into numbers (Yes = 1, No = 0)

Weights (Importance)

• Each input has a weight (W1, W2, …)
• Weight shows the importance of that factor
• More important factor → higher weight

Bias

• Bias is an extra value added to control the decision
• It reflects personal preference (safe or risky decision)

Working (Calculation)

• All inputs are multiplied with their weights
• Then they are added together with bias
• Formula idea: Output = (Inputs × Weights) + Bias

Decision Making

• The result is compared with a threshold (usually 0)
• If result > 0 → Decision = YES
• If result < 0 → Decision = NO