Subjective Type Questions - AI Project Cycle - Artificial Intelligence for Class

Q.1. What do you understand by data exploration? Illustrate the answer with an example.

Ans:

• Data exploration refers to the set of techniques and tools used to examine data so that patterns, anomalies and trends can be identified clearly.
• It commonly uses data visualisation (such as charts and graphs) and simple statistical summaries to understand the shape and structure of the data before modelling.
• For example, when placing an order online, a user looks at product reviews, ratings and feedback left by other customers. Exploring these data helps identify common problems, average satisfaction and likely product quality, which supports better buying decisions.

Q.2. What are the types of Data weak AI systems?

Ans:

The types of data-weak AI systems are as follows:

• Heuristics or rule-based: A user-defined, rule-based system that follows explicit rules to make decisions.
• Brute-force: A method that examines many possible options (for example using a decision tree or search) to find the best choice; classic AI chess programs use this approach to evaluate possible moves.
• Neural networks: Models inspired by the human brain that process data through layers and can improve performance based on input data and feedback.

Q.3. What is the importance of visualising data?

Ans:

• Data visualisation makes complex data easier to understand by showing patterns and trends visually.
• It helps in data discovery and supports evaluation by making relationships and outliers visible.
• Common visual forms include line charts, bar charts, pie charts and scatter plots.
• Visuals are especially useful to compare variables, observe trends over time and communicate findings clearly to stakeholders.

Q.4. Explain some tools used for data visualization.

Ans:

The following are some commonly used tools for data visualisation:

• MS Excel: A widely used spreadsheet application that offers many chart types and basic data analysis features suitable for small to medium datasets.
• Tableau: A tool for creating interactive visualisations and dashboards; it handles large and frequently updated datasets well.
• QlikView: A business intelligence tool that provides strong data-association features and interactive visual analytics.
• FusionCharts: A JavaScript charting library capable of producing many chart types for web-based visualisations.

Q.5. What do you understand by data acquisition?

Ans:

• Data acquisition is the phase in the AI project cycle where data are collected from various sources to train and evaluate models.
• Collected data used to teach the model are called training data, while data used to check model predictions are called testing data.
• Good data acquisition ensures that the model receives sufficient, relevant and high-quality examples for learning.

Q.6. Explain the training data and testing data with an example.

Ans:

• Training data: Historical or existing data used to teach a model the relationship between inputs and outputs.
• Testing data: New or separate data used to evaluate how well the trained model predicts outcomes.
• Example: To predict an employee's salary, past salary records and related features (such as experience, qualification) form the training data. The salary predicted for a new employee, which is then compared with the actual salary, forms the testing data.

Q.7. Justify the sentence - "For any AI project to be efficient, the training data should be authentic and relevant to the problem statement scoped."

Ans:
• Relevant and authentic training data are essential because the model learns patterns from the examples it is given.
• If training data are irrelevant, incomplete or incorrect, the model will learn wrong relationships and make poor predictions on testing data.
• For example, when predicting net run rate in cricket, the model should be trained and tested on batting-related statistics (runs scored, overs faced). If bowling statistics are used instead, the prediction for net run rate will be incorrect.
• Therefore, authentic and relevant training data are crucial for an AI project to perform well.

Q.8. What do you mean by data features?

Ans:

• Data features are the specific pieces of information or attributes that are collected and used by an AI model to make predictions or decisions.
• For example, in a cricket match dataset features could include runs scored by a batsman, runs scored in particular overs, number of wickets fallen, runs conceded by bowlers, and type of dismissal.

Q.9. Mention the ways to collect data.

Ans:

The following ways are commonly used to collect data:

• Surveys
• Web scraping
• Sensors
• Cameras
• Observations
• API
• Call, SMS or email
• Feedback

Q.10. What are some concerns that need to be taken care of while collecting data?

Ans:

• Ensure the data are authentic and accurate.
• Collect data from reliable and verifiable sources.
• Respect intellectual property - prefer open-source or properly licensed data.
• Check for bias, missing values and privacy issues before using the data for modelling.

Q.11. What are the main approaches used for AI modelling?

Ans:

There are two main approaches used for AI modelling:

• Rule-Based Approach
• Learning-Based Approach

Q.12. Explain rule-based approach in detail.

Ans:

• In a rule-based approach, the system follows a set of explicit rules written by developers to produce outputs from given inputs.
• The model applies these rules to data to reach decisions; the relationships are defined by human experts rather than learned from data.
• This approach relies on coding and carefully defined conditions and is suitable when rules are clear and stable.

Q.13. Explain learning-based approach in detail.

Ans:

• In a learning-based approach, the machine is given examples (data) and desired outputs; it then learns patterns or rules from those examples rather than relying on rules written by a developer.
• The developer does not explicitly define the relationship; the model infers it by analysing the data.
• This approach is useful when data are unlabelled or when the patterns are complex and not obvious to humans.
• The machine groups similar data, extracts features and reports the observed trends as output after training.

Q.14. What do you mean by decision tree?

Ans:

• A decision tree is a model used for decision-making and classification that represents choices in a tree-like structure.
• It consists of a topmost root node, internal decision nodes and terminal leaf nodes that show final outcomes.
• Each node is connected by branches that represent the possible options, and the tree is followed from top to bottom to reach a decision.

Q.15. Define the following terms

Ans:

• Root Node: The top node of a decision tree from which all branches originate.
• Splitting: The process of dividing a node into two or more sub-nodes based on a chosen feature.
• Decision or Interior Node: A node where splitting takes place; it leads to one or more sub-nodes.
• Leaf Node or Terminal Node: A node at the bottom of the tree that gives the final outcome or class label.
• Branch or Subtree: A subsection of the decision tree that represents a path of decisions.
• Parent Node and Child Node: A parent node is a node that is split to produce child nodes; a child node is derived from a parent node.

Q.16. What do you mean by the neural network?

Ans:
• A neural network is a computational model inspired by how neurons work in the human brain.
• Also called an Artificial Neural Network (ANN), it processes information through layers of connected nodes and can learn from examples without explicit programming of rules.
• Neural networks can recognise patterns, learn features and make decisions in ways similar to human reasoning for many tasks.

Q.17. Where the neural network used?

Ans:

Neural networks are used in many applications that require recognising patterns from large datasets.

They are commonly used in:

• Voice recognition
• Character recognition
• Signature verification
• Human face recognition

Q.18. How does the neural network work?

Ans:

• The input layer receives the raw data or features.
• Hidden layers process these inputs; each connection has a weight that is adjusted during training.
• Each neuron in a layer applies a simple computation and passes its output to the next layer.
• The output layer produces the final prediction or result based on processed signals from previous layers.
• Training adjusts the weights so that the network's output matches the desired outcomes for the training examples.

Q.19. What are the features of a neural network?

Ans:

• Neural networks are modelled on the human brain's structure.
• They can extract useful information from data without the user writing explicit rules.
• They operate using mathematical functions and adjustable weights in their algorithms.
• They are suitable for large datasets and complex pattern recognition.
• The architecture typically includes an input layer, one or more hidden layers and an output layer.

Q.20. Compare the three types of neural network learning.

Ans:

The comparison is shown in the diagram below:

Q.21. What are the main phases of the AI project cycle?

Ans:

The AI project cycle has the following five main phases:

1. Problem Scoping
2. Data Acquisition
3. Data Exploration
4. Modelling
5. Evaluation

Q.22. What can be done in each phase of the AI project cycle? Write points.

Ans:

The following are typical sub-stages for each phase of the AI project cycle:

1. Problem Scoping
   • Goal setting
   • Identifying the problem
2. Data Acquisition
   • Data collection
   • Defining data requirements
3. Data Exploration
   • Representation of data
   • Visualising data
4. Modelling
   • Model design
   • Feeding data into the model and training
5. Evaluation
   • Testing the project or model
   • Deployment of the model
   • Project review and performance monitoring

Q.23. What do you mean by problem scoping?

Ans:

Problem scoping refers to identifying and defining the problem clearly and setting a vision for solving it.

Problem scoping typically includes:

• Setting the goal
• Identifying and defining the problem
• Brainstorming possible approaches
• Designing, building and testing prototypes
• Showcasing or sharing the solution with stakeholders

Q.24. What is the importance of problem scoping in the AI project cycle?

Ans:

• Problem scoping focuses the project on a specific goal and clarifies what needs to be solved.
• It establishes the sequence: problem definition, brainstorming, design, build, test, and showcase.
• Without clear problem scoping, subsequent stages (data collection, modelling, evaluation) may be misdirected and ineffective.
• Errors in scoping can lead to project failure or delays, so this stage is critical for success.

Q.25. How goal setting can be helpful in problem scoping?

Ans:

• Goal setting narrows the focus to a specific and achievable objective.
• It helps determine what needs to be measured and which data are required.
• Goals can reveal root causes of the problem and reduce the number of possible solutions to consider.
• Well-defined goals make procedures clearer and the project easier to manage.

Q.26. What do you mean by 4Ws canvas?

Ans:

The 4Ws canvas is a simple tool to understand the problem context before starting a project.

It addresses four questions:

• who - People affected directly or indirectly by the problem
• what - The nature of the problem
• where - The context, situation or location
• why - The intended solution and benefits of solving the problem

Q.27. Explain all 4Ws in details for 4Ws canvas.

Ans:

1. Who
   • Under this heading, identify stakeholders and anyone affected by the problem or who will benefit from the solution.
   • Stakeholders include users, customers, managers and others impacted by the outcome.
   • Two important questions to ask are:
     (i) Who are the stakeholders?
     (ii) What do you know about them?
2. What
   • Understand and describe what the problem is.
   • Clarify evidence or observations that show this is a problem worth solving.
3. Where
   • Consider the context, situation or location where the problem occurs and how it affects the solution design.
4. Why
   • Explain why solving this problem is important and what benefits stakeholders will gain from the solution.