Q1. Applications of the Supervised Learning
Select the problem statements where you
can apply supervised algorithms.
1. For an e-commerce website, segmenting the unlabelled customers based on their behaviour from a large dataset.
2. Given data on crop yields over the last 50 years, trying to predict next year's crop yields.
3. Based on data samples of webpages, classifying a webpage whether the content on the web page should be considered "child friendly" or "adult".
4. Given a large dataset of medical records from patients suffering from heart disease, try to learn whether there might be different groups of such patients.
Ans: Correct Answer:
- Given data on crop yields over the last 50 years, trying to predict
next year’s crop yields.
- Based on data samples of webpages, classifying a webpage whether
the content on the web page should be considered “child friendly” or
“adult”.
Explanation:
· Supervised learning is defined by its use of labeled datasets to train algorithms that to classify data or predict outcomes accurately.
· The tasks of predicting crop yields and classifying a web page are based on some data which is used by ML model to learn. The task of predicting crop yields can be done by regression models, while classifying a webpage can be done by classification task.
· Unsupervised learning, uses machine learning algorithms to analyze and cluster unlabeled datasets.
· The task of segmenting customers based on their behaviour is a task of clustering, where ML model finds patterns and makes clusters based on the unlabelled data samples.
· Same goes for grouping patients, based on the medical records of heart disease.
Q2. Applications of Unsupervised learning
What are the
applications of Unsupervised learning ?
1. Grouping of images of different types of animals, based on the image features only
2. Given the dataset with previous 5 years of the prices of the houses. Predicting the house price.
3. Grouping together users with similar viewing patterns in order to recommend similar content.
4. Predicting the class of fruit given we know the type of fruits.
Ans:
Correct Answer:
1.
Grouping of images of different types of animals, based on the image
features only,
- Grouping together users with similar viewing patterns in order to
recommend similar content
Explanation:
· Predicting the house prices and classifing the type of fruit are the use cases of supervised learning.
· In the grouping of different types of animals and recommending similar content, both are unsupervised as there are no class labels given
Q3. Person of Interest
Popular online shopping platforms like Myntra, Flipkart,
and Amazon are trying to group their
clients based on their recent purchases and interests so that they can design a
suitable marketing strategy.
Which of the following Machine
Learning method will be appropriate for doing so?
1. Regression
2. Classification
3. Clustering Methods
4. Recommendation
Ans: Correct Options: 3. Clustering Methods
Explanation:
- Grouping of Customers is done on the basis of common
characteristics so companies can market each group effectively and
appropriately. This grouping of Customers can be effectively done by
Clustering algorithms.
- If it were asked to predict the price of a particular product then
we would use a regression model
- If it were asked to predict whether customers will buy the product
or not then we would use the classification model.
- If we were asked to recommend products to customers based on their
recent purchases then we would use a recommendation model
Q4. Price is right
A firm
is making a shopping mall and they have hired you as a Data Scientist for
evaluating the prices of shops.
Which of the following Machine
Learning task will you use?
1.
Classification
2.
Regression
3.
Clustering Methods
4.
Recommendation
Ans: Correct option: 2. Regression
Explanation:
- The prices of shops are continuous and hence, it is a task of
regression.
- If it were asked to state whether to open a shopping mall or not
then it would be a classification task.
- If it were asked to state whether to group shops based on the type
of items sold then it would be a clustering task.
- It would be a recommendation task if it were asked to suggest a
location for various types of shops.
Q5. Mr. Robot ?
You are
an analyst at SBI Bank and recently discovered customer accounts to be hacked.
So you want to create software
that categorizes each customer's account as hacked or safe.
Which machine learning task will
be helpful in this particular task?
1. Classification
2. Regression
3. Clustering
4. Reinforcement learning
Ans: Correct Answer: 1. Classification
Explanation:
- Since the accounts will be categorized as hacked or un-hacked i.e.
classified into categories/ classes. Hence, it is a classification
problem.
- If it were asked to predict the amount lost by each customer then
it would have been a regression problem.
- If it were asked to group customers by income so as to understand
if there is any pattern behind hacking then it would have been a
clustering problem.
- Reinforcement learning is a machine learning training method based
on rewarding desired behaviors and/or punishing undesired ones and is
hence not applicable.
Q6. Poseidon
Government agencies that predict tsunamis, use
Tsunami-detection buoys to confirm the existence of underwater earthquakes.
These buoys use the height of the
waves and change in water pressure to predict a Tsunami.
Which of the following method
might be used for doing the mentioned task?
1. Regression
2. Classification
3. Clustering
4. Reinforcement learning
Ans: Correct option: 2. Classification
Explanation:
- The task of predicting a Tsunami can yield two classes: Tsunami or
No-Tsunami. And, since we have to predict categorical labels i.e. 2
classes. Hence, it is a Classification task.
- If it were asked to predict the magnitude of an earthquake then it
would have been a regression problem.
- We cannot use the clustering method because it does not help in
predicting tsunamis.
- Reinforcement learning is a machine learning training method based
on rewarding desired behaviors and/or punishing undesired ones and thus
would not help in predicting tsunamis.
Q7. Cells at work
While conducting a Bone-marrow transplant, doctors check
whether the donor’s cells are healthy or not. Several organizations use Machine
learning models to determine healthy cells.
Which of the following machine
learning methods can be used?
1. Regression
2. Classification
3. Reinforcement learning
4. Clustering
Ans: Correct Answer:2 Classification
Explanation:
- Donor's cell can either be 'Healthy' or 'Not Healthy'. Which makes
it a classification task.
- We are not predicting any continuous numerical value and hence
regression method is incorrect
- Grouping cells would not help in predicting Healthy or Not healthy
cells and hence clustering method is incorrect
- Reinforcement learning is a machine learning training method based
on rewarding desired behaviors and/or punishing undesired ones and is not
applicable in this scenario.
Q8. ML Paradigm
A computer program is said to learn from experience E with
respect to some task T and some performance measure P, if its performance on T,
as measured by P, improves with experience E.
Suppose we want to predict the
future prices of cryptocurrency based on some data. According to the definition
given above, What would be a reasonable choice for P?
1. The Price prediction task.
2. Based on the learning over past data, the process of learning to find the patterns within newer data.
3. The accuracy of the algorithm for correctly predicting the future price of crypto-currencies.
4. None of these.
Ans: The correct option is:
3.
The accuracy
of the algorithm correctly predicting the future price of crypto-currencies.
Explanation:
· As per the definition, ‘P’ is some performance
measure that needs to be improved over time. So, the accuracy is one of the way
we can measure the performance of predicting future prices.
·
As per the
definition, ‘T’ is the task, so The price prediction task would actually be a
‘T’ in our case.
·
And, the
process of learning to find the patterns within newer data with past data
refers to experience ‘E’
Q9. Unlabelled
What type of
machine learning algorithm is suitable for unlabeled data?
1. Regression algorithms
2. Clustering algorithms
3. Classification
4. All of the above
Ans: Correct answer: 2. Clustering
algorithms
Explanation:
- Both regression and classification methods require labeled data
i.e. supervised learning.
- However, clustering doesn't require labeled data.
Q10. Penny
for suggestion?
What procedure
or method powers Spotify's suggestion for a new song?
1. Recommendation System
2. Classification
3. Regression
4. None of the above
Ans: Correct answer: 1. Recommendation System
Explanation :
- Classification and regression methods are used in supervised
learning task and hence won't be applicable here.
- The option None of the above is incorrect as
Recommendation System is the correct answer.
Q11. Fly away
Imagine you are working in Vistara. You are given a task
of predicting
the airline fare based on previous trends.
What kind of machine learning
task would this be categorized into?
1. Classification
2. Recommendation
3. Time series forcasting
4. Reinforcement learning
Ans: Correct Answer: 3. Time series forecasting
Explanation:
- Whenever we want to predict values based on previous trends i.e.
time axis is involved, the problem is categorized into "Time series
forecasting"
- Classification is not applicable here as we are predicting the
airline fare based on previous trends.
- Recommendation is not correct as we need to find the airline fare
based on previous trends
- Reinforcement learning is not applicable as reinforcement learning
is a machine learning training method based on rewarding desired behaviors
and/or punishing undesired ones
