Class XII Artificial Intelligence Vocabulary Flashcards

Vocabulary flashcards covering core AI concepts, Data Science methodology, Computer Vision, Big Data, Neural Networks, Generative AI, and Data Storytelling from the CBSE Class XII AI Handbook.

Data Science Methodology

A prescribed sequence of 10 iterative steps put forward by John B. Rollins of IBM Analytics that provides a structured framework for data scientists to design and execute AI projects from problem definition to model deployment and feedback.

Descriptive Analytics

A type of data analytics focused on summarizing past historical data using statistical measures (such as mean, median, mode, variance, and standard deviation) and visual tools like graphs to describe trends and patterns.

Diagnostic Analytics

A type of data analytics that analyzes historical data using techniques like root cause analysis and correlation analysis to understand the underlying causes and factors behind why specific events occurred.

Predictive Analytics

A type of data analytics that applies techniques such as regression, classification, and clustering to historical data to forecast future events, patterns, or behaviors.

Prescriptive Analytics

A type of data analytics that uses optimization, simulation, and decision analysis to recommend specific courses of action to achieve a desired future outcome.

Primary Data Source

An original source of raw, unprocessed, and unbiased data gathered firsthand through direct observation, experiments, surveys, interviews, or IoT sensors.

Secondary Data Source

Pre-existing stored data collected from external or internal sources such as books, journals, transactional databases, web scraping, and public online repositories like UNICEF, WHO, data.gov, or Kaggle.

Feature Engineering

A component of data preparation where raw variables are selected, modified, or combined to create new meaningful features to enhance machine learning model performance.

Train-Test Split

A model validation procedure that divides a dataset into two subsets: a training dataset used to fit the model and a testing dataset used to evaluate model predictions.

K-Fold Cross Validation

A model validation technique where a dataset is split into $k$ equal subsets (folds), iteratively training the model on $k-1$ folds and testing it on the remaining fold until every fold is used as a test set once.

Precision

An evaluation metric for classification tasks measuring the proportion of true positive predictions out of all positive predictions, defined as $\text{Precision} = \frac{\text{TP}}{\text{TP} + \text{FP}}$.

Recall

An evaluation metric for classification tasks measuring the proportion of actual positive instances correctly predicted, defined as $\text{Recall} = \frac{\text{TP}}{\text{TP} + \text{FN}}$.

F1 Score

The harmonic mean balancing Precision and Recall, defined as $\text{F1 Score} = 2 \times \frac{\text{Precision} \times \text{Recall}}{\text{Precision} + \text{Recall}}$.

Accuracy

An evaluation metric measuring the overall proportion of correct predictions made by a model, defined as $\text{Accuracy} = \frac{\text{TP} + \text{TN}}{\text{TP} + \text{TN} + \text{FP} + \text{FN}}$.

Pixel

Short for 'picture element', it is the smallest individual square or component of a digital image that holds a numerical intensity or color value.

Histogram Equalization

An image preprocessing technique that spreads out pixel intensity values evenly to adjust brightness and contrast, enhancing details in very dark or bright regions.

Object Detection

A computer vision task that identifies, locates, and draws labeled bounding boxes around multiple object instances of interest within an image using algorithms like YOLO, SSD, or R-CNN.

Semantic Segmentation

An image segmentation approach that classifies every pixel in an image into specific object classes without distinguishing individual instances of the same class.

Instance Segmentation

An image segmentation approach that classifies every pixel while creating distinct pixel-wise masks for each separate object instance, even if they belong to the same class.

Orange Data Mining Tool

An open-source, component-based visual programming software package used for data visualization, data mining, machine learning, and visual workflow building.

Widgets

Graphical workflow components within the Orange data mining environment designed to execute specific data handling, preprocessing, modeling, visualization, or evaluation tasks.

Big Data

Vast and complex datasets originating from transactional, machine, and social sources that exceed the processing and storage capabilities of traditional software tools.

3Vs Framework

The three foundational characteristics defining Big Data: Volume (scale of data), Velocity (speed of data generation and delivery), and Variety (diversity of data formats).

6Vs Framework

An extended framework for Big Data that adds Veracity (data quality and trustworthiness), Value (business benefits derived), and Variability (data context consistency) to the original 3Vs.

Mining Data Streams

The process of analyzing and extracting real-time patterns, trends, and actionable knowledge from continuous incoming flows of data without storing all data permanently.

Artificial Neural Network (ANN)

A machine learning architecture modeled on human brain biological neurons composed of interconnected node layers that extract features and make decisions automatically.

Activation Function

A mathematical decision-making component inside a artificial neuron (e.g., Sigmoid, Tanh, ReLU) that introduces non-linearities to determine whether a neuron should activate and pass its output signal.

Backpropagation

Short for 'backward propagation of errors', an optimization algorithm used during neural network training to adjust connection weights and biases based on the loss computed in previous iterations.

Perceptron

A fundamental single-layer neural network developed by Frank Rosenblatt in 1958 that uses Threshold Logic Units (TLUs) for binary classification tasks.

Convolutional Neural Network (CNN)

A deep neural network architecture featuring a 3D arrangement of artificial neurons and filter layers, specialized for processing spatial and visual data like digital images.

Recurrent Neural Network (RNN)

A neural network architecture featuring internal feedback loops designed to retain temporal memory and process sequential data types such as text and speech.

Generative AI

A branch of artificial intelligence utilizing deep learning and neural networks to autonomously generate new synthetic data—including text, images, video, and audio—resembling its training datasets.

Generative Adversarial Network (GAN)

A neural network framework consisting of two models—a Generator creating fake data and a Discriminator evaluating authenticity—competing against each other to produce realistic synthetic outputs.

Variational Autoencoder (VAE)

A generative deep learning model comprising an Encoder that compresses input data into a latent space representation and a Decoder that reconstructs the data to generate new synthetic samples.

Large Language Model (LLM)

A specialized deep learning algorithm trained on massive text corpora using Transformer architectures to comprehend, classify, translate, and generate human-like text.

Discriminative Models

Machine learning models trained to learn conditional probability distributions and decision boundaries between dataset classes in order to classify input features correctly.

Data Storytelling

The practice of combining data analysis, contextual narratives, and graphical visuals to effectively communicate complex analytical insights to technical and non-technical stakeholders.

Freytag's Pyramid

A dramatic five-stage narrative structure comprising Introduction, Rising Action, Climax, Falling Action, and Conclusion, adapted to organize data presentations into engaging analytical arcs.