Worksheet Solutions: AI Project Cycle

Multiple Choice Questions (MCQs)

Q1: How many stages are there in the AI Project Cycle?
 a) Three
 b) Four
 c) Five
 d) Six
Ans: (c)

Explanation: The AI Project Cycle is made up of five stages that guide the development of an AI solution: Problem Scoping, Data Acquisition, Data Exploration, Modeling and Evaluation. Treating the cycle as five stages helps teams follow a clear, step-by-step process from defining the problem to checking the model's performance.

Q2: What is the first stage of the AI Project Cycle?
 a) Data Acquisition
 b) Problem Scoping
 c) Modeling
 d) Evaluation
Ans: (b)

Explanation: Problem Scoping is the first stage because it sets the project goal and defines the exact problem to be solved. Clear scoping prevents wasted effort later by identifying what success looks like, who is affected and what constraints apply.

Q3: Which block of the 4Ws Problem Canvas identifies the people affected by the problem?
 a) What
 b) Where
 c) Why
 d) Who
Ans: (d)

Explanation: The 'Who' block identifies stakeholders - the people or groups impacted by the problem or the proposed solution. Knowing the 'Who' helps prioritise needs, collect relevant data and design solutions that truly benefit the target users.

Q4: What is the purpose of the 'What' block in the 4Ws Problem Canvas?
 a) To determine the context of the problem
 b) To identify stakeholders
 c) To determine the nature of the problem and gather evidence
 d) To list benefits of the solution
Ans: (c)

Explanation: The 'What' block focuses on defining the problem itself and gathering evidence that it exists. This may include data, articles or observed facts that show why the problem needs solving and helps set measurable goals.

Q5: Which modeling approach involves feeding the machine data along with predefined rules?
 a) Learning-Based Approach
 b) Rule-Based Approach
 c) Supervised Learning
 d) Unsupervised Learning
Ans: (b)

Explanation: The Rule-Based Approach uses explicit rules written by humans together with input data. It is generally static because the system follows the given rules and does not adapt automatically to new patterns unless the rules are changed.

Q6: What is a key characteristic of neural networks?
 a) They require manual feature extraction
 b) They automatically extract data features
 c) They are limited to small datasets
 d) They do not involve hidden layers
Ans: (b)

Explanation: Neural networks learn useful features from raw data through their hidden layers. This automatic feature extraction reduces the need for manual design of inputs and makes neural networks suitable for complex data such as images and audio.

Q7: Which Python tool is recommended for interactively developing AI-related projects?
 a) Anaconda Prompt
 b) Jupyter Notebook
 c) Virtual Environment
 d) Python Shell
Ans: (b)

Explanation: Jupyter Notebook offers an interactive interface with code cells, rich text and visualisations. It is widely used for experimentation and teaching because students can run code step by step and immediately see outputs and plots.

Q8: What is the purpose of a virtual environment in Python?
 a) To execute Python code directly
 b) To isolate project dependencies
 c) To install Jupyter Notebook
 d) To compile Python code
Ans: (b)

Explanation: A virtual environment creates an isolated workspace for a project so that libraries and Python versions do not conflict with other projects or the system Python. This helps maintain reproducibility and avoids dependency issues.

Q9: Which Python operator is used to calculate the remainder of a division?
 a) /
 b) //
 c) %
 d) 
Ans: (c)

Explanation: The '%' operator is the modulo operator in Python; it returns the remainder after division. For example, 7 % 3 gives 1 because 7 = 2 × 3 + 1.

Q10: What does the 'Why' block of the 4Ws Problem Canvas focus on?
 a) Identifying stakeholders
 b) Determining the context of the problem
 c) Listing benefits of the solution for stakeholders
 d) Gathering evidence for the problem
Ans: (c)

Explanation: The 'Why' block records the benefits that stakeholders and society will gain from solving the problem. This justification helps assess impact and prioritise efforts during project planning.

Fill in the Blanks

Q1: The AI Project Cycle provides a framework to achieve the __________ of an AI project.

Ans: goal

The goal of an AI project defines its primary objectives and desired outcomes, guiding all stages from data collection to evaluation.

Q2: The __________ Canvas helps summarize the key elements of the problem into a single template.

Ans: Problem Statement

The Problem Statement Template condenses the important parts of the 4Ws into a clear summary that can be shared with stakeholders and team members.

Q3: In the __________ stage, data is collected to form the base of the AI project.

Ans: Data Acquisition

The Data Acquisition stage gathers raw information and relevant records that will be used for analysis, model training and testing.

Q4: __________ Learning involves training a model with labeled data.

Ans: Supervised

Supervised Learning trains models using labelled examples so the model can learn to map inputs to known outputs and make accurate predictions on new data.

Q5: The __________ layer of a neural network is responsible for providing the final output to the user.

Ans: output

The output layer produces the final prediction or result after the data has been processed through the hidden layers.

True or False

Q1: The Rule-Based Approach in AI modeling allows the machine to adapt to new data dynamically.

Ans: False

Explanation: Rule-based systems follow explicit rules written by humans. They do not change their behaviour automatically when new data appears unless the rules are updated manually, so they are considered static.

Q2: The 4Ws Problem Canvas includes Who, What, Where, and Why blocks to analyze a problem.

Ans: True

Explanation: The 4Ws Canvas structures problem scoping into four parts - Who, What, Where and Why - which together help teams clearly define the problem and its context.

Q3: Data features refer to the type of data needed to address a problem, such as salary or increment percentage.

Ans: True

Explanation: Data features are the specific attributes collected for modelling, for example salary amount or percentage increase, which a model can use as inputs to make predictions.

Q4: Unsupervised learning models always require labeled data to identify patterns.

Ans: False

Explanation: Unsupervised learning uses unlabeled data and finds structure or groups within it, such as clustering similar items without prior labels.

Q5: Jupyter Notebook can only be used with Python and not with other programming languages.

Ans: False

Explanation: Jupyter supports many languages via kernels, including R, Julia and others, not only Python. This makes it versatile for different programming needs.

Short Ans Questions

Q.1: What is the purpose of Problem Scoping in the AI Project Cycle?
Ans: Problem Scoping sets the goal of an AI project by clearly defining the problem to be solved and identifying the key parameters and constraints that affect it. This clarity helps the team plan data needs and choose suitable modelling approaches.

Q.2: Explain the role of the "Who" block in the 4Ws Problem Canvas.
Ans: The "Who" block identifies stakeholders who are directly or indirectly affected by the problem. It records who faces the issue, who will benefit from a solution and relevant details that help design user-centred interventions.

Q.3: What is the difference between a Rule-Based Approach and a Learning-Based Approach in AI modeling?
Ans: In the Rule-Based Approach, humans provide explicit rules and data, producing a static system that does not change unless rules are updated. The Learning-Based Approach trains models on data so they can learn patterns and adapt to new data, allowing dynamic improvement over time.

Q.4: Why are virtual environments useful when working with Python projects?
Ans: Virtual environments keep each project's libraries and Python version isolated, preventing conflicts between projects and preserving a stable development setup. This ensures that each project's dependencies do not affect the base environment or other projects.  

Q.5: Describe the role of the input layer in a neural network.
Ans: The input layer receives raw data and passes it into the neural network without transforming it. It provides the initial set of values that the hidden layers will process to extract features and build representations.

Long Ans Questions

Q.1: Explain the five stages of the AI Project Cycle.
Ans: The AI Project Cycle consists of five stages:

1. Problem Scoping: Define the problem and set a clear project goal. Use tools such as the 4Ws Problem Canvas to capture who is affected, what the problem is, where it happens and why solving it matters.
2. Data Acquisition: Collect and gather the raw data needed for the project. Identify the relevant data features and ensure data quality and coverage suitable for the task.
3. Data Exploration: Visualise and summarise data to find trends, missing values and relationships. This helps choose appropriate features and informs modelling decisions.
4. Modeling: Build models using rule-based or learning-based approaches. Select algorithms, train models on labelled or unlabelled data and tune parameters to improve performance.
5. Evaluation: Test the trained model on unseen data and measure its effectiveness using metrics such as accuracy, precision, recall and F1 score. Use evaluation results to refine the model or revisit earlier stages if needed.

Q.2: Describe the 4Ws Problem Canvas and how each block contributes to problem scoping.
Ans: The 4Ws Problem Canvas organises problem scoping into four parts:

• Who: Identifies stakeholders and the people affected by the problem, helping to focus the solution on real users.
• What: Defines the nature of the problem and gathers evidence (for example, data or reports) to confirm its existence and scale.
• Where: Describes the context, setting or location where the problem occurs, which influences how a solution should be implemented.
• Why: Lists the benefits of solving the problem for stakeholders and society, providing a rationale and measures of success.
  These blocks collectively provide a structured approach to understanding and summarizing the problem for effective scoping.

Q.3: Discuss the difference between supervised and unsupervised learning models with examples.
Ans: Supervised learning uses labeled data to train models, where each data point is tagged with a label. For example, a dataset with images of apples and bananas labeled as such allows the model to predict whether a new image is an apple or banana. Unsupervised learning uses unlabeled data to identify patterns, such as clustering or dimensionality reduction. For instance, clustering can group unknown data based on patterns, while dimensionality reduction simplifies high-dimensional data like words in NLP. Supervised learning relies on predefined labels, while unsupervised learning discovers patterns independently. 

Q.4: Explain how neural networks work, including the roles of the input, hidden, and output layers.
Ans: Neural networks simulate a network of simple processing units to learn from data:

• Input Layer: Accepts raw input values and passes them into the network.
• Hidden Layers: Perform successive transformations and extract features from the input. Each node applies weights and an activation to combine information from the previous layer.
• Output Layer: Produces the final prediction or result for the user, such as a class label or numerical value.

By training on many examples, the network adjusts its internal weights so that the output matches desired results. The automatic feature extraction in hidden layers makes neural networks effective for complex data.

Q.5: Evaluate the importance of creating a virtual environment for Python projects and how it is done using Anaconda.
Ans: Virtual environments are important because they isolate dependencies for each project, avoiding conflicts between packages or Python versions. Using Anaconda, a virtual environment is created as follows:

1. Open the Anaconda Prompt, which starts in the base environment.
2. Create a new environment specifying a Python version, for example, conda create -n env python=3.7.
3. Confirm installation when prompted (type Y).
4. Activate the environment using conda activate env; the prompt will show the active environment name.
5. Install needed packages, for example conda install ipykernel nb_conda jupyter to enable Jupyter Notebook in that environment.

This procedure gives a clean, reproducible workspace so projects remain stable and easy to manage.