UNIT – 1 CAPSTONE PROJECT

I. MULTIPLE CHOICE QUESTIONS 1 MARK EACH

1. A _____________ is a project where students must research a topic independently to

find a deep understanding of the subject matter.
a)Program b)Application
c)Capstone project d)Software
Answer: c)Capstone project
2. Without any ___________, AI cannot solve a problem.
a) pattern b)Data
c)Explanation d) Questions
Answer: a) pattern
3. Which of these statements are CORRECT related to Loss Function?
1. It’s a way of determining how well a certain algorithm models the data.
2. Loss function learns to increase the prediction error over time with the help of
some ‘optimization /objective function”.
3. Log loss is a type of Classification Loss function.
4. Quantile Loss is a type of Regression Loss.

a. Only 1 b. Only 1,2,3

c. Only 1,3,4 d. All are correct
Answer: c. Only 1,3,4
4. _____________ is a design methodology that provides a solution-based approach to
solving problems.
a)Problem Scoping b)Data frame
c)Design thinking d)Data set
Answer: c)Design thinking
5. Match the following
1. MSE A. patterns
2. Core of every AI Model B. Anomaly Detection
3. Is this unusual? C.MAE
4. Regression D. Regression
5. Train-Test Split E. Sensitive to Outliers
 a. 1->A, 2-B, 3->C, 4->D, 5->E b. 1->D, 2->E, 3->A, 4->B, 5->C
c. 1->E,2->D, 3->C, 4->B , 5->A d. 1->E, 2->A, 3->B, 4->C ,5->D
Answer: 1->E, 2->A, 3->B, 4->C ,5->D
6. Every project starts with ___________
a)Business understanding b)Modelling
c)Testing d)Deployment
Answer: a)Business understanding
7. Given 80% of data is selected for training and remaining 20% for testing, and this process
is carried out for four times and error rate is averaged out, this validation technique can be
called as _______
a) Hold-out b) K-fold cross-validation
c) LOOCV d) Bootstrapping
Answer: b) b) K-fold cross-validation

8. A _________ is a set of historical data in which the outcomes are already known.
a)Training set b)Test set
c)Data requirement d)Model
Answer: a)Training set
9. You can compute the residual in regression by
a) actual y‐coordinate value - predicted y‐coordinate value
b) predicted y‐coordinate value - actual y coordinate value
c) actual y‐coordinate value / predicted y‐coordinate value
d) None
Answer: a) actual y‐coordinate value - predicted y‐
coordinate value
10. ________ is used to evaluate the machine learning model.
a)Train model b) Test Dataset
c)Train data set d) Test model
Answer: b) Test Dataset
11. _________ is an iterative process with a prescribed sequence of steps that are
followed by data scientists to approach a problem and find a solution.
a)Robotic methodology b) Engineering
c)Design thinking d)Data Science methodology
Answer: d) Data Science methodology
12. In problem decomposition:
i. Understand the problem and then restate the problem in your own words
ii. Gather all simple facts to create a complicated piece
iii. Break larger units into simpler ones
iv. Code one small unit at a time
Which of the following is true?
(a) (i) and (ii) b) (i), (iii) and (iv)
c ) (ii) and (iv) d) (i), (ii), (iii), (iv)
Answer: (b) (i), (iii) and (iv)

13. Which of the following is NOT a Classification Loss function?

(a) Log Loss/ (b) Hinge Loss
(c) Exponential loss (d) Mean Absolute Error
Answer: (d) Mean Absolute Error

14. helps us to summarize all the key points into one single outline so that
in future, whenever there is a need to look back at the basis of the problem, we can take
a look at it and understand the key elements of it.
a) 4W Problem canvas b) Problem Statement Template
c) Data Acquisition d) Algorithm
Answer: b) Problem Statement Template

15. Which of these is the code for test data split of 0.33?
(a) x_train,x_test,y_train,y_test=train_test_split (x,y,test_size=0.33)
(b) x_train,x_test,y_train,y_test=train_test_split (x,y,test_size=0.67)
(c) x_train,x_test,y_train,y_test=train_test_split (x,y,test_size=0.5)
(d) x_train,x_test,y_train,y_test=train_test_split (x,y,test_size=1)
Answer: (a) x_train,x_test,y_train,y_test=train_test_split (x,y,test_size=0.33)

Assertion and Reasoning:

Answer the following questions, choosing the correct option out of the following:
a. Both A and R are true and R is the correct explanation of A.
b. Both A and R are true and R is not the correct explanation of A.
c. A is true but R is false.
d. A is false but R is true.

Assertion(A): Nowadays, predictive models are not able to better predict rare events such as
disease or system failure.
Reason(R): Today’s high-performance database analytics enable data scientists to utilize
large or even all of the available data.
Answer: d. A is false but R is true.
Assertion (A) – Consider that the goal of an AI model is to predict an answer such as "yes" or
"no".
Reason (R) – In such a case, predictive modelling can be used.
Answer c. A is true but R is false.
II. Answer the given questions in 20 – 30 words each
(2 marks each)

a) ______ is a methodology which details a solution-based technique to solving problems.

By using the different phases in this methodology, anyone can solve complex problems
that arise around us. Identify these phases and define them.

(i) (ii)

Answer: Methodology – Design Thinking

(i) Empathise – You develop an empathic knowledge of the issue you're attempting
to address. The best way to do this is usually through user research.
(ii) Ideate - You seek out alternate perspectives on the issue and seek out creative
answers to the problem statement you've formulated. Here, brainstorming is
really beneficial.
b) Prerna and her team have developed an AI model which predicts house prices in a
city. The RMSE of her AI model is 500. Do you think that Prerna has developed a
good model? List 2 things that RMSE indicates.
Answer: No RMSE of a good AI model should be less than 180. Things that RMSE
indicates:
• It shows how well the model fits the data.
• It gives the average model prediction error expressed in units of the target
variable.
c) Daksh wants to create an AI model to predict lung cancer. He wants to collect data for
his model. Create a list of questions to decompose this task for Daksh.
Answer:
• What data to collect? (coughing blood, shortness of breath, loss of appetite
etc.)
• What should be the timeline of the data (1 month, 6 month, a year etc.)
• What are the sources of authentic data?
• Can I collect this data from a particular region?
• Which income group to concentrate on? (low, high or both)
d) This method CANNOT be used in cases where small data sets necessitate multiple
setups, and when the data set is unbalanced. Identify the method and define it.
Answer: This method is the ‘Train-Test Split’. It is a procedure that measures the
performance of machine learning algorithms when they need to make predictions on
data that was not used to train the model.
It is a quick and simple process to carry out, and the outcomes let you evaluate the
effectiveness of machine learning algorithms for your particular predictive modelling
problem.
 e) Name the four data collection and cleaning steps in the Data Science Methodology.
Which step takes the longest time? Can the time be reduced? How?
Answer: Data Requirement→Data Collection→Data Understanding→Data
Preparation
Data preparation is usually the most time-consuming procedure in any data science
project. However, some data preparation procedures are common for a variety of
problems. Automating certain data preparation steps in advance can speed up the
process by minimizing preparation time. Today's high-performance, massively
parallel systems and analytics capabilities allow data scientists to prepare data more
easily and quickly.

f) Fill in the blanks: MSE is known as the ______ of the error value, while RMSE is the
________ of errors.
Answer: MSE is known as the Variance of the error value, while RMSE is the
Standard Deviation of errors.

III. Answer the given questions in 80-100 words

(4 marks each)
a) “RMSE gives high weight to large errors.” Do you agree? Why/Why not? Give
reasons.
Answer: The residuals are a measurement of how far the data points are from the
regression line. Residuals are prediction errors, which we can determine by deducting
the predicted value from the actual value. In order to get RMSE, we compute the
square root of the average of the squared residuals. The spread of these residuals is
determined by the RMSE. In other words, it reveals the degree to which the data is
centred around the line of best fit.
The RMSE gives significant errors a vastly higher weight since the errors are squared
before being averaged. Since, these are negatively orientated scores, lower values are
preferable.
b) Define the distance marked as ‘A’ in the graph. Why is it important?

Answer: ‘A’ is called residual. The differences between actual and predicted data
values are known as residuals in statistical or machine learning models. They serve as
a diagnostic tool for evaluating a model's quality. They are also called errors.
Importance of residual:
• Residuals are crucial in assessing a model's quality.
 • When all of the residuals are 0, that means that the model makes accurate
predictions. The model is considered biased when the average residual is not
0. (i.e., consistently over- or under-predicting).
• When residuals show patterns, the model is most likely qualitatively
inaccurate because it is unable to account for some aspect of the data.
c) Sudha is studying the Data Science Methodology. Answer the following questions for
Sudha.
i) Why does data need to be split?
ii) Differentiate between the Training Data and Test Data.
Answer:
i. To assess how effectively our machine learning model works, we must divide
a dataset into train and test sets. The statistics of the train set are known and
are utilised to fit the model. The test data set, which is the second set, is
utilised just for predictions.
ii. A collection of data used to fit the model is called the training dataset. This
dataset is used to train the model. The model observes and learns this data.
An accurate assessment of a final model fit is provided by the test dataset,
which is a subset of the training dataset. Testing of the AI model is conducted
after model training.

d) Calculate the MSE and RMSE values of the following data

Predicted value 14 19 17 13 12 7 24 23 17 18

Observed Value 17 18 18 15 18 11 20 18 13 19

Answer:
Residual
(observed - Squared
Predicted Observed predicted) Residuals
14 17 3 9
19 18 -1 1
17 18 1 1
13 15 2 4
12 18 6 36
7 11 4 16
24 20 -4 16
23 18 -5 25
17 13 -4 16
18 19 1 1
Total 125
MSE 12.5
RMSE 3.54
e) “Regression functions predict a quantity, and classification functions predict a label.”
What are these functions collectively called? Define it and also give 2 examples from
each category.
Answer: These are called loss functions. A loss function evaluates the accuracy with
which your prediction model can forecast the desired result (or value). The learning
problem is converted into an optimization problem, a loss function is defined, and the
algorithm is then optimised to minimise the loss function.
Examples – Regression Loss Function – RMSE, MAPE
Classification Loss Function – Log loss, Focal Loss
f) Neha is studying linear regression analysis between 2 variables. She finds that the
residuals are quite large. Answer the following questions for Smitha:
i) What are residuals
ii) What is their role in model evaluation?
iii) What does their value imply?
Answer:
i. A residual in a linear regression is the difference between the dependent
variable's predicted value and its actual observed value.
ii. By measuring how well the model matches the data, it is used to assess the
model's effectiveness.
iii. The residuals should be small and uniformly spaced around the mean if the
model successfully fits the data. The model may not be a good match for the
data if the residuals are large and not equally distributed, and it may need to be
improved or revised.
g) “Cross Validation is doing train-test split several times.” Do you agree? Differentiate
between train test split approach and Cross validation technique with visualizations.
Answer: Yes, the statement is true.
Train Test Split Approach Cross Validation Technique
Dataset is divided into two parts – At the core it is doing train-test split
Training Set and the unseen Test Set several times.
Data points used for training are not Every data point could potentially be in
used for testing. either the testing or the training set.
Better for large datasets. More robust for small datasets.

Image Source : https://www.baeldung.com/cs/cross-validation-k-fold-loo

 UNIT – 2 MODEL LIFECYCLE

I. MULTIPLE CHOICE QUESTIONS 1 MARK EACH

1. During the initial data collection phase, data scientists identify available data sources
relevant to the problem area. Identify the data sources.
i) Structured
ii) Unstructured
iii) Semi-structured
iv) Re-structured

(a) All the four sources (b) Only 1 and 2

(c ) Only 1 and 3 (d) Only 1,2,3
Answer: (d) Only 1, 2, 3
2. Which of these is not an AI platform?
(a) Microsoft Azure AI Platform (b) Google Cloud AI Platform
(c ) IBM Watson Developer Platform (d) HDFC AI Platform
Answer (d) HDFC AI Platform
3. In this phase, we define the project’s strategic business objectives and desired
outcomes, align all stakeholder’s expectations as well as establish success metrics.
Identify this phase of the AI Model Life Cycle.
(a) Design (b) Scoping
(c ) Evaluation (d) Data collection
Answer: (b) Scoping
4. Which of the following are the objective of the testing team in AI modelling?
i) Model Validation ii) Minimizing bias
iii) Security compliance iv) Understanding data
(a) Only i, ii, iii (b) Only ii, iii, iv
(c ) Only i, iii, iv (d) Only i, ii, iv
Answer: (a) Only i, ii, iii
5. Choose the five stages of AI project cycle in correct order
(a) Evaluation->Problem Scoping->Data Exploration->Data Acquisition->Modelling
(b) Problem Scoping->Data Exploration->Data Acquisition->Evaluation->Modelling
(c) Problem Scoping->Data Acquisition->Data Exploration->Modelling->Evaluation
(d) Data Acquisition->Problem Scoping->Data Exploration->Modelling->Evaluation
Answer: (c ) Problem Scoping->Data Acquisition->Data Exploration->Modelling->
Evaluation
6. Models are not built once, deployed and forgotten about. They are constantly refined
and adapted to changing situations. This is done using __________, which tells about
the model’s effectiveness. This can come in the form of response rates to a
promotional campaign.
(a) Scoping (b) Feedback
(c) Evaluation (d) Design
Answer: (c ) Feedback
7. Name 2 sources of authentic data.
Answer: Government websites, sensors
 8. Radha is working on an imbalanced dataset. For such a dataset which metric
should she use to evaluate the AI model.
(a) Accuracy
(b) Precision
(c) F1 Score
(d) Pearson’s Coefficient
Answer: (c) F1 Score
9. Assertion (A): During “Exploratory Data Analysis”, the initial insights gained help to
get an understanding of the data and later on, help in algorithm selection, metrics
choice, etc.
Reason (R): AI Project Life Cycle is an iterative process.
Select the correct option:
a. Both A and R are true and R is the correct explanation of A
b. Both A and R are true and R is not the correct explanation of A
c. A is true but R is false
d. A is False but R is true
Answer: b. Both A and R are true and R is not the correct explanation of A

II. Answer the given questions in 20 – 30 words each

(2 marks each)

1. Differentiate between Data Acquisition and Data Understanding stages.

Answer:
Data Acquisition Data Understanding
Data is collected from credible Data is cleaned to locate missing values,
sources. eliminate worthless data (erroneous samples and
outliers).
No statistical analysis is Basic statistical analysis such as drawing graphs
performed in this stage. and comparing different properties of the data set
are carried out. The initial insights gained help to
get an understanding of the data and later on, help
in algorithm selection, metrics choice, etc. This
complete procedure is called “Exploratory Data
Analysis”.

2. “ML engineers go through multiple models to determine the best model

configuration.” During which stage of the AI Project Life Cycle does this happen.
Briefly explain this stage.
Answer: Modelling is the process through which several models based on graphical
data can be constructed and even tested for advantages and disadvantages. ML
engineers go through multiple models to determine the best model configuration.
Hence, the design phase is an iterative process. Hyperparameter fine-tuning provided
by most ML frameworks helps to narrow down the number of feasible solutions.
3. In this stage, the AI model is made available for use to the customers. Identify this stage
and state its relevance.
 Answer: This stage is Deployment. This is the final phase of the machine learning
life cycle. The deployed model’s performance is monitored to ensure that it continues
to function at the level required by the business. Sometimes, retraining of the model is
also done to make it more robust.
4. In which stage of the AI Project Life Cycle does visualization of data occur?
Answer: This occurs in the Data Exploration stage. Here, basic statistical analysis
such as drawing graphs and comparing different properties of the data set are carried
out. The initial insights gained help to get an understanding of the data and later on,
help in algorithm selection, metrics choice, etc. It also helps to understand which
elements are more essential and what the overall trend of the data is.

III. Answer the given questions in 80-100 words

(4 marks each)
1. Suppose you are creating an AI model which is a version of Google Translate. What
all questions will you consider in the Scoping stage?
Answer:
o What are the commercial constraints?
o Your model must be capable of comprehending text data. The end result should
be linguistically correct as feasible.
o Minor mistakes will be acceptable due to language complexities.
o For a better user experience, the result should be shown in milliseconds.

2. A company has created an AI model to predict amount of nitrogen in soil. The model
was tested extensively and then deployed to customer. Now, however, the model is
not giving reliable insights. What could be the reason for this? Give a solution to this
problem also.
Answer: The model is overfitting. A model is overfitted when it is trained with a lot
of data. As the model now remembers the entire dataset, it should be retrained with a
completely new set of data and evaluated using techniques like k-fold cross validation.

3. Manoj is part of the testing team for developing an AI model. He is told “The testing
team should put the AI and ML algorithms through rigorous testing while maintaining
model validity and keeping successful learning ability, and algorithm efficacy in mind.”
What steps should Manoj now take?
Answer:
• Data validation should be performed on test data to check for bias.
• Test data should include all relevant subsets of training data, i.e., the data you
will use for training the AI system.
• Manoj’s team must create test suites that help validate the ML models.
• The AI model can be tested using cross validation technique. This technique
will give a better idea of model performance.

4. Reeta is working as a data scientist. She is tasked with studying the learning rate of
the model. She finds that the learning rate is quite low. What should she do now?
What category of parameter does learning rate belong to?
Answer: Learning rate is a hyperparameter. Since the learning rate is low, the model
is taking too much time for training/learning. Hence this hyperparameter needs to be
tweaked for optimal model performance.
 5. Consider the following:
“Is this soap dispenser racist?” This question became an internet sensation. In
a video at a Marriott hotel in Atlanta in 2015, an automatic soap dispenser was
shown unable to detect a black customer’s hand. What is this called in AI
terminology? When an AI model behaves like this, which stage should you go back
to? Give reasons.
Answer: The above episode depicts AI bias. When an AI model behaves like this, we
should go back to data collection stage. The dataset should have more samples from
diverse population groups. If the AI model is trained on the right data, it will give
better and accurate results.

UNIT – 3 STORYTELLING THROUGH DATA

I. MULTIPLE CHOICE QUESTIONS 1 MARK EACH

1. is the process of using fact and narrative to communicate something to your

audience.
i) Letter Writing (b) Storytelling
(c ) AI Project Cycle (d) Problem Statement Template
Answer: (b) Storytelling
2. _____ serves as the foundation of your data story.
i) Character (b) Plot
(c ) Data (d) Theme
Answer: (c ) Data
3. ____ are charts, graphs, images or videos.
i) Theme (b) Conflict
(c ) Data (d) Visualisations
Answer: (d) Visualisations
4. Data storytelling uses a _________ approach to delivering data insights.
i) narrative (b) structured
(c ) unstructured (d) realistic
Answer: (b) structured
5. ___________ is the first step involved in Data Storytelling.
i) Understanding the audience (b) Visualisations
(c ) Narrative (d) Plot
Answer: (a) Understanding the audience
6. Fill in the blanks:
i) When narrative is combined with data, it helps to _____ to your audience
what’s happening in the data and why a specific insight is essential.
ii) When visuals are applied to data, they can _____ the audience to insights that
they wouldn’t infer without charts or graphs.
iii) When narrative and visuals are combined, they can ____ or even entertain an
audience.

(a) enlighten, engage, explain

(b) explain, enlighten, engage
(c) explain, engage, enlighten
(d) All explain
 Answer: (b) explain, enlighten, engage
7. Assertion (A): Stories that combine statistics and analytics are not persuasive.
Reason (R): When we visualize data through charts and graphs, they can enlighten the
audience to see the data in a different perspective and help them to analyse and make
the right decisions at the right time.
Select the correct option:
a. Both A and R are true and R is the correct explanation of A
b. Both A and R are true and R is not the correct explanation of A
c. A is true but R is false
d. A is False but R is true
Answer: d. A is false but R is true.
II. Answer the given questions in 20 – 30 words each
(2 marks each)
1. Name 2 components of a data story.
Answer: Visuals and Data
2. Namrata is writing steps to create an effective data story. However, she has got all the
points mixed up. Rearrange the points for Namrata in the correct sequence.
i. Include authentic data
ii. Examine the relationships in the data—pinpoint the significant data
iii. The context must be relevant
iv. Use a simple narrative surrounded by conflict
v. Add in compelling visuals
Answer:
• The context must be relevant
• Include authentic data
• Add in compelling visuals
• Examine the relationships in the data—pinpoint the significant data
• Use a simple narrative surrounded by conflict

3. Manisha is studying data storytelling. List 2 ethics that Manisha should follow while
creating a data story.
Answer:
• Respect privacy.
• Avoid making assumptions about the experiences of others.

4. What is the purpose of visualization in data storytelling? Name few visualization

tools.
Answer: Data visualization is defined as a visual representation of information. Visuals
play an important role in telling a story and communicating important information. Data
stories link the audience with the data. Data visualizations support and enhance data
stories, helping to communicate your findings elegantly and efficiently.
Visualization Tools – Charts, Infographic

5. Manju is part of a team developing an AI model. What will data stories created by
Manju and her team convey in the Data Exploration step?
Answer: Data Exploration is the first step in developing a model. In this step, a data
analyst uses visual exploration to understand what is in a dataset and the characteristics
 of the data, rather than through traditional data management systems. So, the data
stories will reflect the data characteristics.

III. Answer the given questions in 80-100 words

(4 marks each)
1. What does the term “Understanding the Audience” mean in context to data
storytelling?
Answer: When creating and presenting your data stories, keep your audience in
mind at all times. Your intended audience must find the story you wish to convey
entertaining and relevant for it to have the impact you desire. Consider the following
as you develop your datastory:
• Is this story appropriate for my audience?
• Does it offer them the necessary knowledge or resolve an issue they care
about?
• Have they previously heard this story?
Your audience's age, gender, occupation, and level of subject matter expertise will
all have an impact on how they interpret and react to your stories.
2. Explain how the three elements of a data story can influence change.
Answer: Data storytelling uses a structured approach to deliver data insights - data,
visuals, and narrative. When a narrative is backed by data, it helps to explain to the
audience what is happening in the data and why a specific insight was developed.
When visuals, such as charts or graphs, are applied to data, they can enlighten the
audience to insights that they would not have noticed otherwise. Finally, when
narrative and images are combined, an audience can be engaged or even entertained.
When suitable visuals and narrative are combined with the correct data, you have a
data story that has the potential to impact and drive change.

3. Consider the following visual. Create an effective data story using this visual.

Image Source: Election Commission of India

Answer:
Indian society is evolving.
Women are voting more than
ever. This is a positive change!

2014
Female Turnout – 65%
Male turnout – 67%

4. Identify and define the following icons related to data storytelling. How can these
elements bring about change?
a. Data b. Narrative c. Visuals

Answer:
a) Data- Facts and information.
b) Narrative – the skill or process of telling a story
c) Visuals - a graphical representation such as a graph, chart, or other
presentation.
When the proper visuals and narrative are combined with the correct data, you
have a data story that has the potential to impact and drive change.
5. Anita is studying about data stories. She creates a data story on women education
in rural areas in India. However, her teacher tells her that her story lacks structure.
What structure should Anita follow? Explain.
Answer: Anita should structure her story as follows:

Setup Conflict Resolution

• Setup - the context of the story which will include the problem, the objective, the data
sources, and the audience.
• Conflict - the difficulties and obstacles that the characters of the story have to
overcome, such as information gaps, constraints, risks, or trade-offs.
• Resolution - the outcome/solution are given in the resolution. This can also include
any insights, suggestions, advantages, or next actions.
6. Answer the following briefly:
i. Name two digital tools which help to visualise data.
Answer: Tableau, Google Charts
ii. Which one out of the following is a better visual? Give reasons.
(a)

(b) Wordpress Installs by language

Image source: https://wptavern.com/wordpress-stats-page-redesigned-adds-

new-data-on-installs-by-langauge
Answer: The first image is a better visual because it is easy to understand and
comprehend. The second graph has too many partitions which are difficult to
navigate.
7. Create an effective data story from the following data.
Semmelweis was a Hungarian physician. In 1844, Semmelweis worked as an
assistant at an obstetrics clinic where many women suffered childbed fever – a
fever caused by an infection post-delivery. During his stint at the clinic,
Semmelweis was intrigued by a peculiar trend. The clinics operated by male
physicians and medical students had significantly higher mortality rates from
childbed fever as compared to clinics operated by midwives. As a result, he
introduced hand-washing. Child mortality rate which was 12.2% dropped to
2.2%. However, some doctors refused to listen to him. Then the mortality rate
reached 3.5%. When the mortality rate touched 4.9%, Semmelweis was thrown
out of the community. (Source : Internet)
 Answer:

8. Create an effective data story with the following data

(Image Source: https://www.statista.com/chart/21780/deadliest-cyclones-india-

bangladesh/)
 Answer:

(Please Note: Data Story can vary as per student’s understanding and creativity)

************************