AI Project Cycle– Class 9 AI (417)
AI Project Cycle
Class 9 Artificial Intelligence, Unit 1.2 — complete notes, 4Ws canvas, MCQs, fill in the blanks, true or false, Q&A and a solved AI Project Cycle case study.
1.2 AI Project Cycle — notes
AI Project Cycle is the cyclical process followed to complete an AI project, taking us through the different steps involved in it.
- It helps us create better AI projects easily.
- It helps us create AI projects faster.
- It helps us understand the process of building an AI project.
- It has six stages: Problem Scoping, Data Acquisition, Data Exploration, Modelling, Evaluation and Deployment.
This entire cycle is illustrated in the book through the CottonAce case, an app that helps cotton farmers detect Pink Bollworm infestation using AI.
1.2.1 Problem Scoping
Problem Scoping means identifying a problem that exists around us and having a vision to solve it using AI.
- It is the first stage of the AI Project Cycle.
- It begins with selecting a broad theme, such as Agriculture or Health, and narrowing it down to a specific topic.
- Various problems are listed under the chosen topic, and one problem is selected as the goal of the project.
- The 4Ws Problem Canvas and the Problem Statement Template are used to define the problem clearly.
4Ws Problem Canvas
The 4Ws Problem Canvas is a tool that helps in identifying the key elements related to a problem by looking at it from four different angles: Who, What, Where and Why.
- Who? This block helps in analysing the people who are getting affected directly or indirectly by the problem; these people are called stakeholders, and they are the ones who face the problem and would benefit from its solution.
- What? This block helps in determining the exact nature of the problem and requires gathering evidence, such as newspaper articles, media reports or announcements, to prove that the problem actually exists.
- Where? This block focuses on the context, situation and location in which the problem arises and is most prominent.
- Why? This block requires thinking about the benefits that the stakeholders and society would get once an ideal solution to the problem is implemented.
Problem Statement Template
The Problem Statement Template summarises all the key points from the 4Ws Canvas into a single template, so that the basis of the problem can be looked back at whenever needed.
General format, as given in the book:
| Template | Details | Corresponding W |
|---|---|---|
| Our | [stakeholders] | Who |
| has a problem that | [issue, problem, need] | What |
| when / while | [context, situation] | Where |
| An ideal solution would | [benefit of solution for them] | Why |
Filled example for the CottonAce (Pest Management) case study:
| Template | Details | Corresponding W |
|---|---|---|
| Our | Farmers | Who |
| has a problem that | Cotton crops got infected with pest – Pink Bollworm | What |
| when / while | On the crops in the field | Where |
| An ideal solution would | Create an AI-enabled app that aids farmers in determining the correct amount of pesticides, knowing the right time to spray pesticides, increasing production, and increasing the farmers' profit share | Why |
1.2.2 Data Acquisition
Data Acquisition is the process of collecting authentic and relevant data required to solve the scoped problem and train the AI model.
- Data can be collected through surveys, web scraping, sensors, cameras and observations.
- The most reliable data comes from open-source government portals, such as data.gov.in and india.gov.in, since random internet sources can be unreliable and private data cannot be used without permission.
- The specific type of data to be collected, such as images or numeric records, is called a data feature.
Training Data is the data on which the AI model is trained before it is put to use. For example, an employee's past five years of salary records can be used to teach the model how salaries usually increase.
Testing Data is the new data that is fed into the already trained model to check the accuracy of its predictions. For example, the trained model can be used to predict the same employee's salary for the next year, and this prediction is then compared with the actual figure.
System Map
A System Map is a diagram that shows the relationship between different elements of a system using directional arrows.
- A '+' sign on the arrow shows a direct relationship, meaning that if one factor increases, the other factor connected to it also increases.
- A '–' sign on the arrow shows an inverse relationship, meaning that if one factor increases, the other factor connected to it decreases.
- A longer arrow between two elements represents a time delay between the cause and its effect.
1.2.3 Data Exploration
Data Exploration is the process of arranging and visualising the acquired data to identify trends, patterns and relationships within it.
- It is done before modelling so that the correct model can be chosen for the problem.
- It helps in quickly identifying trends and patterns in large amounts of data.
- It helps in communicating the findings of the data effectively to others through visual means.
- Common visualisation tools include bar graphs, line graphs, pie charts, scatter plots and histograms.
1.2.4 Modelling
Modelling is the process of building an algorithm or a model, using the explored data, that can perform the required task by mimicking human intelligence.
Artificial Intelligence (AI) is any technique that enables a computer to mimic human intelligence and behaviour.
Machine Learning (ML) is a subset of Artificial Intelligence in which machines improve their performance at a task through experience gained from data, without being explicitly programmed for every situation.
Deep Learning (DL) is a subset of Machine Learning in which the software trains itself using a very large amount of data, applying several layers of machine learning algorithms to learn on its own.
- AI, ML and DL are related in a nested manner: ML is a part of AI, and DL is a part of ML, as shown in the diagram below.
Rule-Based Approach is a modelling approach in which the developer defines a fixed set of rules, and the machine follows these rules while performing the given task.
- The machine's learning is static in this approach, meaning it does not learn or improve on its own.
- Once the rules are set, the machine will not change its output even if the input data pattern changes.
- It is suitable for problems where the relationship between inputs and outputs is well known and fixed.
Learning-Based Approach is a modelling approach in which the machine is trained on a large set of data along with the desired output, and it learns to design its own rules based on this data.
- The machine's learning is dynamic in this approach, meaning it keeps adapting itself according to the data it is exposed to.
- It is suitable for problems where the data patterns are complex or keep changing over time.
1.2.5 Evaluation
Evaluation is the process of testing the reliability of an AI model by feeding it a set of test data and comparing its predicted results with the actual, real results.
- Evaluating a model on the same data that was used to train it is called overfitting, and it should be avoided since the model only memorises the training data instead of learning to generalise.
- True Positive (TP): the model predicts "Yes" and the actual result is also "Yes".
- True Negative (TN): the model predicts "No" and the actual result is also "No".
- False Positive (FP): the model predicts "Yes" but the actual result is "No".
- False Negative (FN): the model predicts "No" but the actual result is "Yes".
- Different algorithms can be evaluated and compared using such results to select the model that performs the best.
1.2.6 Deployment
Deployment is the final stage of the AI Project Cycle in which the tested and evaluated AI model or solution is implemented in a real-world scenario for actual use by the stakeholders.
- It involves testing and validating the final model before it goes live.
- It involves integrating the AI model with the existing system, such as a mobile app or website, so that end users can access it.
- It involves regularly monitoring and maintaining the deployed model to ensure it keeps working correctly over time.
Multiple Choice Questions
1.Which one of the following is the second stage of the AI project cycle?
Show answer
2.Which of the following comes under Problem Scoping?
Show answer
3.Which of the following is not a valid method for Data Acquisition?
Show answer
4.If an arrow goes from X to Y with a – (minus) sign in a system map, it means that:
Show answer
5.Which of the following is not a part of the 4Ws Problem Canvas?
Show answer
6.The process of breaking down a big problem into a series of simple steps is known as:
Show answer
7.The primary purpose of data exploration in the AI project cycle is:
Show answer
8.Does modelling mean creating an AI model?
Show answer
9.Can we use AI on mobile phones?
Show answer
Fill in the Blanks
1.Rearrange the steps of the AI project cycle in the correct order: (a) Data Acquisition (b) Problem Scoping (c) Modelling (d) Data Exploration (e) Deployment (f) Evaluation
Show answer
2.Identify A, B and C in the AI–ML–DL diagram.
Show answer
True or False
1.Deployment is the final stage in the AI project cycle where the AI model or solution is implemented in a real-world scenario.
Show answer
Question & Answer
1.What are the various stages of the AI Project Cycle? Explain each with an example.
Show answer
- Problem Scoping – identifying the problem to be solved, for example, pest infestation damaging cotton crops.
- Data Acquisition – collecting relevant data, for example, images of pests captured by farmers.
- Data Exploration – organising and visualising the collected data to find patterns.
- Modelling – building an AI algorithm, for example, an image-recognition model that identifies pests.
- Evaluation – testing the model's accuracy, for example, the CottonAce model gave 70% correct results initially.
- Deployment – implementing the final solution, for example, installing the CottonAce app on farmers' mobile phones.
2.How is an AI project different from an IT project?
Show answer
3.Explain the 4Ws problem canvas in problem scoping.
Show answer
- Who? identifies the stakeholders, that is, the people who face the problem and would benefit from its solution.
- What? identifies the exact nature of the problem along with evidence that proves it exists.
- Where? identifies the context, situation and location in which the problem occurs.
- Why? identifies the benefits that the stakeholders and society would get from an ideal solution.
4.Why is there a need to use a Problem Statement Template during problem scoping?
Show answer
5.What is Problem Scoping? What are the steps of Problem Scoping?
Show answer
6.Who are the stakeholders in the problem scoping stage?
Show answer
7.How will you differentiate between Training Data and Testing Data? Elaborate with examples.
Show answer
8.Name various methods for collecting data. For each method, name at least one project in which you may use that method of data collection.
Show answer
- Surveys – used in a project to collect farmer feedback on pesticide use.
- Web scraping – used in a project that gathers product reviews from e-commerce sites.
- Sensors – used in a project that records the number of pests caught in a field trap.
- Government open-data portals – used in a project on predicting disease outbreaks using health statistics.
9.What must you keep in mind while collecting data so that it is useful?
Show answer
- The data collected must be authentic and accurate.
- The data must be relevant to the problem statement that has been scoped.
- The data must not be someone's private property, and it should be collected ethically from reliable sources.
10.Draw a system map for enabling farmers to take their produce to the market for sale.
Show answer
11.Name a few government websites from where you can get open-source data.
Show answer
12.What is the significance of Data Exploration after acquiring data? Explain with examples.
Show answer
13.What is the relevance of Data Visualization in AI?
Show answer
14.List any five graphs used for data visualization.
Show answer
15.How is Data Exploration different from Data Acquisition?
Show answer
16.Use an example to explain a Data Visualization technique.
Show answer
17.What is Artificial Intelligence? Give a day-to-day example.
Show answer
18.How is Machine Learning related to Artificial Intelligence?
Show answer
19.Compare and contrast Rule-based and Learning-based approaches in AI modelling, indicating when each may be used.
Show answer
- In a Rule-Based approach, the developer defines fixed rules, and the machine simply follows them; it is useful when the relationship between input and output is well known and unlikely to change.
- In a Learning-Based approach, the machine is trained on data and designs its own rules, adapting to new or changing data; it is useful when the patterns in the data are complex or keep evolving.
20.What is Evaluation?
Show answer
21.What are the various model evaluation techniques?
Show answer
- Comparing the model's predictions against the actual results using True Positive, True Negative, False Positive and False Negative outcomes.
- Using metrics such as ROC to compare the accuracy of different algorithms and selecting the best one.
22.Why is model evaluation important in AI projects?
Show answer
23.What do you understand by the terms True Positive and False Positive?
Show answer
24.Case-study question: A hospital uses an AI model to predict whether a patient has a disease. Out of 100 patients tested, the model gave the following results: 40 patients were correctly predicted as having the disease, 45 patients were correctly predicted as not having the disease, 10 patients were wrongly predicted as having the disease when they did not, and 5 patients were wrongly predicted as not having the disease when they actually did. Identify the True Positive, True Negative, False Positive and False Negative values from this case.
Show answer
- True Positive (TP) = 40 – patients correctly predicted as having the disease.
- True Negative (TN) = 45 – patients correctly predicted as not having the disease.
- False Positive (FP) = 10 – patients wrongly predicted as having the disease.
- False Negative (FN) = 5 – patients wrongly predicted as not having the disease.
25.What is deployment in the context of an AI project cycle?
Show answer
26.Why is deployment an important phase in the AI project cycle?
Show answer
27.What are some common challenges in deploying AI models?
Show answer
- Integrating the model smoothly with existing systems, such as mobile apps or websites.
- Ensuring the model works reliably under real-world, uncontrolled conditions.
- Carrying out ongoing monitoring and maintenance of the model after it has been deployed.
Case Study (4 Marks)
Question: Map the steps of the "Preventable Blindness" case study, that is, AI-based detection of Diabetic Retinopathy at Aravind Eye Hospital, to the AI Project Cycle.
Show answer
1. Problem Scoping – Stage: identifying the problem and setting a clear goal to solve it. In this case: the goal is to prevent blindness caused by Diabetic Retinopathy, which occurs due to delayed diagnosis from a lack of qualified doctors and slow report generation.
2. Data Acquisition – Stage: collecting relevant and authentic data needed to train the AI model. In this case: digital images of patients' retinas are collected from many clinics using retinal cameras operated by trained technicians.
3. Data Exploration – Stage: organising and validating the collected data to find patterns before modelling. In this case: the retinal images are validated and made sense of, so that consistent, usable data is available to build the diagnostic model.
4. Modelling – Stage: building an AI algorithm or model that can perform the required task using the explored data. In this case: an AI model is created, in partnership with Google, that can correctly diagnose Diabetic Retinopathy when given a retinal image as input.
5. Evaluation – Stage: testing the model's accuracy and fine-tuning it for better performance. In this case: the model is tested for accuracy and further fine-tuned; it achieved 98.6% accuracy, on par with specialist eye doctors.
6. Deployment – Stage: implementing the tested solution in a real-world setting for actual use. In this case: the AI tool is deployed in clinics and vision centers across rural and remote parts of the country, currently used in 71 vision centers in rural Tamil Nadu, enabling any technician, even without a skilled doctor, to screen patients early.