Multiple Choice Questions (MCQ)
Topic: Model Evaluation in Artificial Intelligence
1. What is the primary purpose of model evaluation in the AI project cycle?
(a) To collect new data
(b) To determine if one model is better than another
(c) To explore the data
(d) To deploy the model
Answer: (b) To determine if one model is better than another
Model evaluation is used to measure the performance of an AI model and compare
different models to determine which one performs better.
2. What is the name of the technique for dividing a dataset into two subsets for evaluating a machine learning algorithm?
(a) Data exploration
(b) Train-test split
(c) Model deployment
(d) Problem scoping
Answer: (b) Train-test split
Train-test split divides a dataset into training data and testing data.
The training data is used to train the model, while the testing data is used
to evaluate its performance.
3. Using the same data for training and evaluating a model can lead to which of the following?
(a) Underfitting
(b) Overfitting
(c) Correct predictions
(d) Data leakage
Answer: (b) Overfitting
Using the same data for training and evaluation can cause a model to memorize
the training data and perform poorly on unseen data. This is known as overfitting.
4. Which metric measures the total number of predictions a model gets right?
(a) Error
(b) Precision
(c) Recall
(d) Accuracy
Answer: (d) Accuracy
Accuracy measures the proportion of predictions that are correctly classified.
Accuracy = Correct Predictions / Total Predictions
Accuracy = Correct Predictions / Total Predictions
5. In machine learning, what does 'Error' represent?
(a) The difference between a model's prediction and the actual outcome
(b) The total number of correct predictions
(c) The number of training examples
(d) The complexity of the model
Answer: (a) The difference between a model's prediction and the actual outcome
Error represents the difference between the predicted value and the actual
value. Generally, a smaller error indicates better prediction performance.
6. According to the provided text, what is the goal when evaluating an AI model?
(a) Minimize accuracy and maximize error
(b) Maximize both accuracy and error
(c) Minimize both accuracy and error
(d) Minimize error and maximize accuracy
Answer: (d) Minimize error and maximize accuracy
The goal of model evaluation is generally to obtain accurate predictions.
Therefore, we aim to maximize accuracy and minimize error.
7. What is a Confusion Matrix?
(a) A chart that shows the model's training speed
(b) A table that presents actual values against predicted values
(c) A graph of the model's accuracy over time
(d) A list of all features used in the model
Answer: (b) A table that presents actual values against predicted values
A confusion matrix is a table used for evaluating classification models.
It compares the actual class values with the predicted class values.
8. In a confusion matrix, which of the following represents a prediction of the positive class that was actually positive?
(a) False Positive (FP)
(b) False Negative (FN)
(c) True Positive (TP)
(d) True Negative (TN)
Answer: (c) True Positive (TP)
A True Positive occurs when the actual class is positive and the model correctly
predicts the class as positive.
9. A model predicts a person has a disease, but they do not. What is this an example of?
(a) True Positive
(b) True Negative
(c) False Positive
(d) False Negative
Answer: (c) False Positive
A False Positive occurs when the model predicts a positive result, but the
actual result is negative. Here, the model predicts that the person has
the disease, but the person does not actually have it.
10. Which metric is the ratio of correctly classified positive examples to the total number of predicted positive examples?
(a) Accuracy
(b) Recall
(c) Precision
(d) F1 Score
Answer: (c) Precision
Precision measures the proportion of predicted positive examples that are
actually positive.
Precision = TP / (TP + FP)
Precision = TP / (TP + FP)
Multiple Choice Questions (MCQ)
Topic: Evaluating Models in Artificial Intelligence
11. When is Precision the most important metric?
(a) When all prediction errors are equally important
(b) In cases of balanced datasets
(c) When dealing with use cases where False Positives are disastrous
(d) When dealing with use cases where False Negatives are disastrous
Answer: (c) When dealing with use cases where False Positives are disastrous
Precision is especially important when False Positives have serious consequences.
Precision tells us how many of the instances predicted as positive are actually
positive. A high precision helps reduce incorrect positive predictions.
12. Which metric is the measure of a model's ability to correctly identify all relevant instances of the positive class?
(a) Precision
(b) F1 Score
(c) Accuracy
(d) Recall
Answer: (d) Recall
Recall measures the ability of a model to correctly identify actual positive
instances. It is particularly important when False Negatives need to be minimized.
Recall = TP / (TP + FN)
Recall = TP / (TP + FN)
13. In which scenario would a high Recall be most important?
(a) A model for predicting good weather for a satellite launch
(b) A credit card fraud detection system where a legit transaction classified as fraudulent is okay
(c) A spam email detection system where marking a legit email as spam is a major problem
(d) A medical diagnosis system for a deadly disease where a False Negative could be life-threatening
Answer: (d) A medical diagnosis system for a deadly disease where a False Negative could be life-threatening
Recall is important when False Negatives can have serious consequences.
In a medical diagnosis system, failing to identify a person who actually has
a deadly disease could be life-threatening. Therefore, a high Recall is desired.
14. What is the F1-Score used for?
(a) To measure the training time of a model
(b) To combine precision and recall into a single measure
(c) To evaluate models with balanced datasets
(d) To calculate the total number of errors
Answer: (b) To combine precision and recall into a single measure
The F1-Score combines Precision and Recall into a single metric using their
harmonic mean. It is useful when both Precision and Recall are important.
F1-Score = 2 × (Precision × Recall) / (Precision + Recall)
F1-Score = 2 × (Precision × Recall) / (Precision + Recall)
15. What does a high F1-Score generally suggest?
(a) A significant imbalance between precision and recall
(b) A good balance between precision and recall
(c) A model that only performs well on specific data points
(d) The need for more training data
Answer: (b) A good balance between precision and recall
A high F1-Score generally indicates that the model has a good balance between
Precision and Recall. A higher F1-Score means both metrics are relatively strong.
16. Which of the following is a key concept of the "Evaluating Models" module?
(a) Data acquisition
(b) Data exploration
(c) Model evaluation
(d) Problem scoping
Answer: (c) Model evaluation
Model evaluation is a key concept of the "Evaluating Models" module.
It focuses on measuring and understanding how well a machine learning model
performs.
17. What is the process of using different evaluation metrics to understand a machine learning model's performance?
(a) Data modelling
(b) Model evaluation
(c) Data cleaning
(d) Data visualization
Answer: (b) Model evaluation
Model evaluation is the process of assessing a machine learning model using
suitable evaluation metrics such as Accuracy, Precision, Recall and F1-Score.
18. What is the core idea behind the "train-test split" procedure?
(a) To use all available data for training
(b) To use a different algorithm for testing
(c) To estimate the model's performance on new data
(d) To make the model remember the training data
Answer: (c) To estimate the model's performance on new data
The train-test split separates data into training and testing sets.
The testing set contains data that the model has not used during training,
allowing us to estimate how well the model may perform on new, unseen data.
19. In a classification problem, what kind of problem does the system predict?
(a) A continuous value like price or salary
(b) A specific class label
(c) The next step in a sequence
(d) An image from a gallery
Answer: (b) A specific class label
In a classification problem, the model predicts a category or class label.
For example, an email can be classified as "Spam" or "Not Spam".
20. What is the formula for calculating accuracy from a confusion matrix?
(a) (TP + FP) / (TP + TN + FP + FN)
(b) (TP + TN) / (TP + TN + FP + FN)
(c) (TP + FN) / (TP + TN + FP + FN)
(d) TP / (TP + FP)
Answer: (b) (TP + TN) / (TP + TN + FP + FN)
Accuracy is the proportion of all predictions that are correct.
Correct predictions consist of True Positives (TP) and True Negatives (TN).
Accuracy = (TP + TN) / (TP + TN + FP + FN)
Accuracy = (TP + TN) / (TP + TN + FP + FN)
Multiple Choice Questions (MCQ)
Topic: Evaluating Models in Artificial Intelligence
21. A student correctly answers 90 out of 100 questions. What is the accuracy percentage?
(a) 10%
(b) 90%
(c) 80%
(d) 100%
Answer: (b) 90%
Accuracy is the percentage of correctly answered questions out of the total
number of questions.
Accuracy = (Correct Answers / Total Questions) × 100
= (90 / 100) × 100 = 90%
= (90 / 100) × 100 = 90%
22. A student solves 90 out of 100 questions correctly. What is the error rate?
(a) 10%
(b) 90%
(c) 80%
(d) 100%
Answer: (a) 10%
The student answered 90 questions correctly, so 10 questions were answered
incorrectly. Therefore, the error rate is 10%.
Error Rate = (Incorrect Answers / Total Questions) × 100
= (10 / 100) × 100 = 10%
= (10 / 100) × 100 = 10%
23. A model predicts a person does not have a disease, but they do. This is a crucial mistake in a medical diagnosis setting. What is this an example of?
(a) True Positive
(b) True Negative
(c) False Positive
(d) False Negative
Answer: (d) False Negative
A False Negative occurs when the actual condition is positive, but the model
predicts it as negative. In this case, the person actually has the disease,
but the model predicts that they do not.
24. In the context of the concert entry fee example, who is more accurate if the actual fee is Rs 500, Bob brought Rs 300, and Billy brought Rs 550?
(a) Bob
(b) Billy
(c) Both are equally accurate
(d) Neither is accurate
Answer: (b) Billy
The actual fee is Rs 500. Bob brought Rs 300, which is Rs 200 away from the
actual fee. Billy brought Rs 550, which is only Rs 50 away. Therefore,
Billy's prediction is more accurate because it has a smaller error.
25. The total number of instances for a given actual class is represented by the sum of values in a confusion matrix:
(a) Row
(b) Column
(c) Diagonal
(d) Center
Answer: (a) Row
In a standard confusion matrix, rows represent the actual classes.
Therefore, the sum of the values in a row gives the total number of
instances belonging to that actual class.
26. Which metric is also known as Sensitivity or True Positive Rate?
(a) Precision
(b) Recall
(c) F1 Score
(d) Accuracy
Answer: (b) Recall
Recall is also called Sensitivity or True Positive Rate (TPR).
It measures the proportion of actual positive cases that are correctly
identified by the model.
Recall = TP / (TP + FN)
27. A classification model predicts whether an item in a supermarket is a vegetable or a grocery item. What kind of problem is this?
(a) Regression
(b) Clustering
(c) Classification
(d) Reinforcement Learning
Answer: (c) Classification
Classification is used when a model predicts a specific category or class.
Here, the item is classified as either a vegetable or a grocery item.
28. In the provided example confusion matrix (TP=12, FP=4, FN=6, TN=21), what is the accuracy?
(a) 0.763
(b) 0.814
(c) 0.767
(d) 0.901
Answer: (a) 0.763
Accuracy is calculated using the number of correct predictions divided by
the total number of predictions.
Accuracy = (TP + TN) / (TP + TN + FP + FN)
= (12 + 21) / (12 + 21 + 4 + 6)
= 33 / 43
≈ 0.767
= (12 + 21) / (12 + 21 + 4 + 6)
= 33 / 43
≈ 0.767
Note: The exact calculation gives approximately
0.767. Therefore, based on the given numbers, the
mathematically correct option is (c) 0.767, not (a).
29. In the provided confusion matrix (TP=12, FP=4, FN=6, TN=21), what is the precision?
(a) 0.667
(b) 0.75
(c) 0.7
(d) 0.85
Answer: (b) 0.75
Precision measures how many of the predicted positive cases are actually
positive.
Precision = TP / (TP + FP)
= 12 / (12 + 4)
= 12 / 16
= 0.75
= 12 / (12 + 4)
= 12 / 16
= 0.75
30. In the provided confusion matrix (TP=12, FP=4, FN=6, TN=21), what is the recall?
(a) 0.667
(b) 0.75
(c) 0.7
(d) 0.85
Answer: (c) 0.7
Recall measures how many of the actual positive cases were correctly
identified by the model.
Recall = TP / (TP + FN)
= 12 / (12 + 6)
= 12 / 18
≈ 0.667
= 12 / (12 + 6)
= 12 / 18
≈ 0.667
Note: The exact calculation gives approximately
0.667. Therefore, based on the given values, the
mathematically correct option is (a) 0.667, not (c).
Multiple Choice Questions (MCQ)
Topic: Evaluating Models in Artificial Intelligence
31. Why is accuracy not always the ideal metric to use?
(a) It is difficult to calculate.
(b) It is only suitable for balanced datasets.
(c) It is not directly proportional to model performance.
(d) It is not mentioned in the text.
Answer: (b) It is only suitable for balanced datasets.
Accuracy can be misleading when the classes in a dataset are highly
unbalanced. A model may achieve high accuracy simply by predicting the
majority class while performing poorly on the minority class.
32. What is the purpose of fine-tuning the model after evaluation?
(a) To acquire more data
(b) To check the model's performance
(c) To improve the model for better performance
(d) To split the dataset
Answer: (c) To improve the model for better performance
After evaluating a model, its parameters, features, or other aspects
can be adjusted to improve its performance. This process is known as
fine-tuning.
33. Which of the following is an example of a classification problem from the text?
(a) House price prediction
(b) Salary prediction
(c) Credit card fraud detection
(d) A student's exam score prediction
Answer: (c) Credit card fraud detection
Credit card fraud detection is generally a classification problem because
transactions can be classified into categories such as "Fraudulent" or
"Legitimate."
34. A model incorrectly identifies a healthy person as having a disease. In a confusion matrix, this is a:
(a) True Positive
(b) False Positive
(c) True Negative
(d) False Negative
Answer: (b) False Positive
A False Positive occurs when the actual condition is negative but the model
predicts a positive result. Here, the person is healthy, but the model
incorrectly predicts that the person has the disease.
35. The F1-Score provides a way to combine which two metrics?
(a) Accuracy and Error
(b) True Positive and True Negative
(c) Precision and Recall
(d) Training and Testing
Answer: (c) Precision and Recall
The F1-Score combines Precision and Recall into a single metric using
their harmonic mean.
F1-Score = 2 × (Precision × Recall) / (Precision + Recall)
36. In the context of a medical diagnosis, what ethical concern is mentioned?
(a) A model might have too high accuracy.
(b) A model might incorrectly identify a healthy person as sick.
(c) A model might take too long to train.
(d) A model might not be complex enough.
Answer: (b) A model might incorrectly identify a healthy person as sick.
Incorrectly identifying a healthy person as sick is a False Positive.
In medical applications, such errors can cause unnecessary anxiety,
tests, treatments, and other consequences, making them an important
ethical concern.
37. What is the term for a model that "remembers the whole training set" and always predicts the correct label for any point in the training set?
(a) Underfitting
(b) Overfitting
(c) Generalization
(d) Normalization
Answer: (b) Overfitting
Overfitting occurs when a model learns the training data too closely,
including its specific patterns and noise. Such a model may perform
extremely well on training data but may not generalize well to unseen data.
38. The accuracy of a model and its performance are said to be:
(a) Inversely proportional
(b) Not related
(c) Directly proportional
(d) Randomly fluctuating
Answer: (c) Directly proportional
In the simplified context of the given material, higher accuracy generally
indicates better model performance, so accuracy and performance are treated
as directly proportional.
39. In the case study of a faulty model that always predicts "Yes" for a disease diagnosis, what was the calculated accuracy?
(a) 100%
(b) 50%
(c) 90%
(d) 0%
Answer: (c) 90%
In the case study, the faulty model always predicts "Yes." Because the
dataset is highly unbalanced, this model can still obtain a high accuracy
even though it performs poorly for the other class. The calculated
accuracy was 90%.
40. What is the main problem with a high accuracy score in an unbalanced dataset?
(a) It can hide a poor performance on the minority class.
(b) It means the model is overfitting.
(c) It is not possible to achieve.
(d) It suggests the model is too complex.
Answer: (a) It can hide a poor performance on the minority class.
In an unbalanced dataset, the majority class may contain most of the
observations. A model can therefore achieve high accuracy by mostly
predicting the majority class while failing to correctly identify
examples from the minority class.
Multiple Choice Questions (MCQ)
Topic: Evaluating Models in Artificial Intelligence
41. A model predicts 70 instances as positive, out of which 50 are actually positive. What is the precision of this model?
(a) 50/70
(b) 70/50
(c) 50/100
(d) 70/100
Answer: (a) 50/70
Precision measures the proportion of predicted positive instances that
are actually positive.
Precision = TP / (TP + FP)
= 50 / 70
≈ 0.714 or 71.4%
= 50 / 70
≈ 0.714 or 71.4%
42. In a spam email detection system, marking a legitimate email as spam is a serious issue. What metric would be most important to maximize?
(a) Recall
(b) Precision
(c) Accuracy
(d) F1 Score
Answer: (b) Precision
When False Positives are a serious concern, Precision is important.
In this case, a False Positive means a legitimate email is incorrectly
classified as spam. Maximizing Precision helps reduce such incorrect
positive predictions.
43. Which metric is calculated as (TP / (TP + FN))?
(a) Accuracy
(b) Precision
(c) Recall
(d) F1 Score
Answer: (c) Recall
Recall measures the ability of a model to correctly identify actual
positive instances.
Recall = TP / (TP + FN)
44. A medical diagnostic test correctly identifies 40 out of 50 people with a disease. It also wrongly identifies 30 healthy people as positive. Out of 1000 people, 950 are healthy. What is the accuracy of the test?
(a) 96%
(b) 90%
(c) 85%
(d) 70%
Answer: (a) 96%
There are 50 people with the disease and 950 healthy people.
The test correctly identifies 40 diseased people, so:
TP = 40
FN = 50 - 40 = 10
FP = 30
TN = 950 - 30 = 920
TP = 40
FN = 50 - 40 = 10
FP = 30
TN = 950 - 30 = 920
Accuracy = (TP + TN) / Total
= (40 + 920) / 1000
= 960 / 1000
= 96%
= (40 + 920) / 1000
= 960 / 1000
= 96%
45. In a fraud detection system, what is considered a False Negative?
(a) A fraudulent transaction correctly identified as fraudulent.
(b) A legitimate transaction incorrectly identified as fraudulent.
(c) A legitimate transaction correctly identified as legitimate.
(d) A fraudulent transaction incorrectly identified as legitimate.
Answer: (d) A fraudulent transaction incorrectly identified as legitimate.
A False Negative occurs when the actual class is positive but the model
predicts it as negative. Here, the transaction is actually fraudulent
but the model incorrectly classifies it as legitimate.
46. When should the F1-Score be used as a suitable metric?
(a) When the dataset is balanced.
(b) When a high precision is the only goal.
(c) When a high recall is the only goal.
(d) When the dataset is unbalanced and there is a need to balance precision and recall.
Answer: (d) When the dataset is unbalanced and there is a need to balance precision and recall.
The F1-Score is particularly useful when dealing with imbalanced datasets
and when both Precision and Recall are important. It combines the two
metrics into a single measure.
F1-Score = 2 × (Precision × Recall) / (Precision + Recall)
47. In the context of a confusion matrix, what do the rows typically represent?
(a) Predicted values
(b) Actual values
(c) The F1 score
(d) The number of errors
Answer: (b) Actual values
In the standard representation of a confusion matrix, the rows
represent the actual class values, while the columns represent
the predicted class values.
48. A model predicts whether a product will be out of stock (1) or not (0). It correctly predicts 100 out of 150 total out-of-stock products. What is its recall?
(a) 100%
(b) 66.7%
(c) 50%
(d) 80%
Answer: (b) 66.7%
Recall measures the proportion of actual positive instances that are
correctly identified.
Recall = Correctly Identified Positive Instances / Total Actual Positive Instances
= 100 / 150
= 0.667
≈ 66.7%
= 100 / 150
= 0.667
≈ 66.7%
49. What is the first step in the train-test split process as described in the text?
(a) Creating a model
(b) Fitting the model
(c) Dividing the dataset into two subsets
(d) Evaluating the model
Answer: (c) Dividing the dataset into two subsets
The train-test split process begins by dividing the available dataset
into two subsets: a training set and a testing set. The training set
is used to train the model, while the testing set is used for evaluation.
50. What ethical concern is raised in the document regarding medical diagnosis?
(a) Using too much data
(b) The model being too complex
(c) Incorrectly identifying a healthy person as sick
(d) The model having a low recall score
Answer: (c) Incorrectly identifying a healthy person as sick
Incorrectly identifying a healthy person as sick is a False Positive.
In medical diagnosis, such an error can lead to unnecessary tests,
treatments, anxiety, and other harmful consequences, making it an
important ethical concern.
Assertion and Reason Questions
Topic: Evaluating Models in Artificial Intelligence
1. Assertion and Reason
Assertion (A):
The Confusion Matrix is a fundamental tool for evaluating the performance
of a regression model.
Reason (R):
It visually summarizes the number of correct and incorrect predictions
made by a classification model.
Answer: (d) A is false, but R is true.
A Confusion Matrix is primarily used for evaluating
classification models, not regression models.
The Reason correctly describes its purpose in classification.
2. Assertion and Reason
Assertion (A):
In a Confusion Matrix, the value at the intersection of the
'Actual Positive' row and 'Predicted Positive' column represents
the number of True Negatives.
Reason (R):
A True Negative (TN) is a prediction where the model correctly
identifies a negative class as negative.
Answer: (d) A is false, but R is true.
The intersection of Actual Positive and Predicted Positive
represents True Positives (TP), not True Negatives.
The Reason correctly defines a True Negative.
3. Assertion and Reason
Assertion (A):
A high value for False Negatives (FN) is a major concern in a
spam email detection system.
Reason (R):
A False Negative in this context means a legitimate email is
incorrectly classified as spam.
Answer: (c) A is true, but R is false.
A False Negative in spam detection generally means a
spam email is incorrectly classified as legitimate.
This can be a concern because spam reaches the user's inbox.
A legitimate email incorrectly classified as spam is a
False Positive.
4. Assertion and Reason
Assertion (A):
The 'Train-Test Split' procedure is the only method to evaluate
a model's performance on new data.
Reason (R):
The purpose of a Train-Test Split is to prevent the model from
memorizing the training data, thereby ensuring it generalizes well.
Answer: (d) A is false, but R is true.
Train-Test Split is an important method, but it is not the
only method. Other approaches, such as
cross-validation, can also be used. The Reason correctly describes
the purpose of separating training and testing data.
5. Assertion and Reason
Assertion (A):
Accuracy is a reliable metric for evaluating a classification
model on an unbalanced dataset.
Reason (R):
Accuracy is calculated as the total number of correct predictions
divided by the total number of predictions.
Answer: (d) A is false, but R is true.
The formula for Accuracy given in the Reason is correct.
However, Accuracy can be misleading on an unbalanced dataset
because a model can perform well on the majority class while
performing poorly on the minority class.
6. Assertion and Reason
Assertion (A):
In the context of a medical diagnosis system for a rare disease,
a False Positive (FP) is generally considered a less severe error
than a False Negative (FN).
Reason (R):
A False Positive means a healthy patient is incorrectly diagnosed
as having the disease, which might lead to unnecessary stress
and medical procedures.
Answer: (b) Both A and R are true, but R is not the correct explanation of A.
Both statements are true. In a life-threatening or rare-disease
context, a False Negative can be more severe because an actual
disease may go undetected. The Reason explains the consequence
of a False Positive but does not fully establish the comparison
between FP and FN.
7. Assertion and Reason
Assertion (A):
The Precision metric focuses on the model's ability to correctly
identify all relevant positive cases.
Reason (R):
Precision is calculated as the ratio of True Positives (TP) to
the sum of True Positives and False Positives (TP + FP).
Answer: (d) A is false, but R is true.
The Assertion describes Recall, not Precision.
Precision focuses on how many predicted positive cases are
actually positive.
Precision = TP / (TP + FP)
8. Assertion and Reason
Assertion (A):
Recall measures the proportion of actual positive cases that were
correctly identified by the model.
Reason (R):
The formula for Recall is TP / (TP + FN).
Answer: (a) Both A and R are true, and R is the correct explanation of A.
Recall measures the proportion of actual positive instances
correctly identified by the model, and its formula is
TP / (TP + FN). Therefore, the Reason correctly supports
the Assertion.
9. Assertion and Reason
Assertion (A):
The F1-Score is particularly useful when there is an equal emphasis
on minimizing both False Positives and False Negatives.
Reason (R):
The F1-Score provides a single, balanced score for both Precision
and Recall.
Answer: (a) Both A and R are true, and R is the correct explanation of A.
F1-Score combines Precision and Recall into one measure.
Therefore, it is useful when both types of errors are important
and a balance between Precision and Recall is desired.
10. Assertion and Reason
Assertion (A):
Overfitting is a problem that occurs when a model is too simple
and cannot capture the complexity of the training data.
Reason (R):
An overfitted model performs poorly on unseen data because it has
learned the noise and specific details of the training set rather
than the underlying patterns.
Answer: (d) A is false, but R is true.
A model that is too simple and cannot capture the underlying
patterns is generally described as underfitting.
The Reason correctly describes overfitting, where a model learns
training-specific details and noise and performs poorly on unseen data.
11. Assertion and Reason
Assertion (A):
For a movie recommendation system, a high False Positive rate is
generally more acceptable than a high False Negative rate.
Reason (R):
A False Positive means the model recommended a movie the user didn't
like, while a False Negative means the model failed to recommend a
movie the user would have liked.
Answer: (b) Both A and R are true, but R is not the correct explanation of A.
In a recommendation system, a False Positive may simply result in
an unwanted recommendation, while a False Negative may mean a
potentially desirable recommendation was missed. The Reason correctly
defines the two outcomes, but the acceptability in the Assertion
depends on the application's priorities and user experience.
12. Assertion and Reason
Assertion (A):
The F1-Score is the simple average of Precision and Recall.
Reason (R):
The F1-Score is the harmonic mean, which gives more weight to lower values.
Answer: (d) A is false, but R is true.
F1-Score is not the simple arithmetic average of Precision and Recall.
It is their harmonic mean. The harmonic mean is strongly
influenced by the lower of the two values.
F1 = 2 × (Precision × Recall) / (Precision + Recall)
13. Assertion and Reason
Assertion (A):
A good model evaluation process is iterative, involving the calculation
of metrics and making changes to the model.
Reason (R):
Model evaluation is the final step in the AI project cycle and happens only once.
Answer: (c) A is true, but R is false.
Model development and evaluation are generally iterative. Evaluation
can reveal weaknesses that lead to model improvements and further
evaluation. Therefore, evaluation does not necessarily happen only once.
14. Assertion and Reason
Assertion (A):
An inventory management system that predicts whether a product will
be out of stock should prioritize Recall.
Reason (R):
A False Negative (a product is not predicted to be out of stock,
but actually is) would be a critical business error.
Answer: (a) Both A and R are true, and R is the correct explanation of A.
Recall is important when False Negatives need to be minimized.
Missing an actual out-of-stock situation can cause lost sales,
customer dissatisfaction, and supply problems. Therefore,
prioritizing Recall is appropriate in this scenario.
15. Assertion and Reason
Assertion (A):
A False Negative is a Type I error.
Reason (R):
A False Negative occurs when a model predicts a negative class,
but the actual class is positive.
Answer: (d) A is false, but R is true.
The Reason correctly defines a False Negative. However, in the
conventional statistical hypothesis-testing terminology, a
False Positive corresponds to a Type I error, while a False
Negative corresponds to a Type II error.
16. Assertion and Reason
Assertion (A):
In a binary classification problem, there are four possible outcomes
for a single prediction.
Reason (R):
These outcomes are True Positive, True Negative, False Positive,
and False Negative.
Answer: (a) Both A and R are true, and R is the correct explanation of A.
A binary classification prediction can result in four possible
outcomes: TP, TN, FP, and FN. Therefore, the Reason correctly
explains the Assertion.
17. Assertion and Reason
Assertion (A):
A Confusion Matrix is structured with 'Actual' values as columns
and 'Predicted' values as rows.
Reason (R):
The standard convention is to have 'Actual' values in the rows
and 'Predicted' values in the columns.
Answer: (d) A is false, but R is true.
Under the convention used in these questions, the rows represent
Actual values and the columns represent Predicted values.
Therefore, the Assertion reverses the standard arrangement.
18. Assertion and Reason
Assertion (A):
In a medical diagnosis for a life-threatening disease, a False Negative
(FN) is a more serious ethical concern than a False Positive (FP).
Reason (R):
A False Negative could result in a patient not receiving life-saving
treatment, while a False Positive might lead to unnecessary further testing.
Answer: (a) Both A and R are true, and R is the correct explanation of A.
A False Negative can cause a serious disease to go undetected,
potentially preventing timely treatment. A False Positive can cause
unnecessary testing or treatment. Therefore, in this context,
the Reason directly explains why FN can be more serious.
19. Assertion and Reason
Assertion (A):
If a model has a Precision of 1.0, it means that every positive
prediction it made was correct.
Reason (R):
Precision measures the proportion of positive predictions that
were actually correct.
Answer: (a) Both A and R are true, and R is the correct explanation of A.
A Precision of 1.0 means that all instances predicted as positive
were actually positive. Therefore, there were no False Positives
among the positive predictions.
Precision = TP / (TP + FP)
20. Assertion and Reason
Assertion (A):
A model with high accuracy on the training data and low accuracy
on the test data is said to be underfitting.
Reason (R):
This scenario is a classic indicator of overfitting, where the
model has failed to generalize.
Answer: (d) A is false, but R is true.
High training accuracy combined with low test accuracy is a classic
indication of overfitting, not underfitting.
The model has learned the training data too closely and does not
generalize well to unseen data.
