ROC curves and AUC quiz

Question 1

What does ROC stand for in the context of classification model evaluation?

Accepted Answer

Receiver Operating Characteristic

Answer

Rate of Classification

Answer

Regression Output Curve

Answer

Recall-Oriented Calculation

Question 2

What two quantities are plotted on the axes of an ROC curve?

Accepted Answer

TPR on y-axis, FPR on x-axis

Answer

Precision on y-axis, Recall on x-axis

Answer

Accuracy on y-axis, Threshold on x-axis

Answer

FPR on y-axis, TPR on x-axis

Question 3

How is the True Positive Rate (TPR) calculated from a confusion matrix?

Accepted Answer

TP / (TP + FN)

Answer

TP / (TP + FP)

Answer

TN / (TN + FP)

Answer

TP / (TP + TN)

Question 4

How is the False Positive Rate (FPR) calculated from a confusion matrix?

Accepted Answer

FP / (FP + TN)

Answer

FP / (FP + TP)

Answer

FN / (FN + TP)

Answer

FP / (TN + FN)

Question 5

What does AUC stand for, and what is its value range?

Accepted Answer

Area Under the Curve; ranges 0 to 1

Answer

Average Under Classification; ranges 0 to 100

Answer

Accuracy Under Curve; ranges -1 to 1

Answer

Area Used in Computation; ranges 0 to 0.5

Question 6

What AUC value corresponds to a classifier that makes random guesses?

Accepted Answer

0.5

Answer

0.0

Answer

1.0

Answer

0.25

Question 7

In the ROC plot, what does the diagonal line from (0,0) to (1,1) represent?

Accepted Answer

A random classifier with no discriminative power

Answer

A perfect classifier

Answer

The decision threshold of 0.5

Answer

The precision-recall tradeoff

Question 8

Which corner of the ROC plot represents the ideal classifier point?

Accepted Answer

Top-left (0, 1)

Answer

Bottom-right (1, 0)

Answer

Top-right (1, 1)

Answer

Bottom-left (0, 0)

Question 9

How is an ROC curve generated from a classifier's predicted probabilities?

Accepted Answer

By varying the decision threshold and computing TPR/FPR at each

Answer

By plotting accuracy at a single fixed threshold

Answer

By averaging precision over all classes

Answer

By sorting features by importance

Question 10

What is a key probabilistic interpretation of the AUC value?

Accepted Answer

Probability a random positive scores higher than a random negative

Answer

The average accuracy across thresholds

Answer

The probability of a correct prediction

Answer

The fraction of correctly classified samples

Question 11

Why is AUC considered threshold-independent, unlike accuracy?

Accepted Answer

It aggregates performance over all thresholds

Answer

It only uses the 0.5 threshold

Answer

It ignores false positives

Answer

It is computed from accuracy alone

Question 12

Why is AUC often preferred over accuracy for imbalanced datasets?

Accepted Answer

Accuracy is inflated by predicting the majority class

Answer

AUC requires balanced data to work

Answer

Accuracy cannot be computed on imbalanced data

Answer

AUC always equals accuracy

Question 13

What is the relationship between specificity and the False Positive Rate?

Accepted Answer

FPR = 1 - specificity

Answer

FPR = specificity

Answer

FPR = 1 + specificity

Answer

FPR = specificity / 2

Question 14

What happens to TPR and FPR as the decision threshold is lowered toward zero?

Accepted Answer

Both TPR and FPR approach 1

Answer

Both approach 0

Answer

TPR rises while FPR falls

Answer

TPR falls while FPR rises

Question 15

Which scikit-learn function computes the AUC score directly from labels and probability scores?

Accepted Answer

roc_auc_score

Answer

accuracy_score

Answer

confusion_matrix

Answer

precision_score

Question 16

What does the scikit-learn roc_curve function return?

Accepted Answer

fpr, tpr, thresholds

Answer

precision, recall, f1

Answer

accuracy, loss, auc

Answer

tp, fp, tn

Question 17

If model A has AUC 0.92 and model B has AUC 0.78 on the same test set, which generally has better discriminative ability?

Accepted Answer

Model A

Answer

Model B

Answer

Both are equal

Answer

Cannot tell from AUC

Question 18

What does an AUC value below 0.5 indicate about a classifier?

Accepted Answer

Worse than random; predictions may be inverted

Answer

A perfect classifier

Answer

Balanced performance

Answer

Identical to AUC 1.0

Question 19

How is the area under the ROC curve typically computed numerically from discrete (FPR, TPR) points?

Accepted Answer

Trapezoidal rule summing areas between points

Answer

Simple averaging of TPR values

Answer

Counting points above the diagonal

Answer

Multiplying max TPR by max FPR

Question 20

How does the ROC curve of a perfect classifier look?

Accepted Answer

Goes to (0,1) then to (1,1), hugging top-left

Answer

A straight diagonal line

Answer

A flat horizontal line

Answer

A downward curve toward (1,0)

Question 21

What is the main difference between an ROC curve and a Precision-Recall curve?

Accepted Answer

ROC uses TPR vs FPR; PR uses precision vs recall

Answer

They are identical

Answer

ROC ignores positives entirely

Answer

PR uses TPR vs FPR

Question 22

Does the AUC of an ROC curve depend on the class balance of the test set?

Accepted Answer

Largely insensitive because TPR and FPR are per-class rates

Answer

Highly dependent on the positive-to-negative ratio

Answer

Only works on perfectly balanced data

Answer

Always increases with more positives

Question 23

What is Youden's J statistic and how does it relate to the ROC curve?

Accepted Answer

J = TPR - FPR; its max gives an optimal threshold

Answer

J = precision + recall

Answer

J is the area under the curve

Answer

J = accuracy at threshold 0.5

Question 24

In a medical screening test where missing a disease is very costly, how should the decision threshold be adjusted?

Accepted Answer

Lower the threshold to raise sensitivity

Answer

Raise the threshold to reduce sensitivity

Answer

Keep it fixed at 0.5 always

Answer

Set it to 1.0

Question 25

For a multi-class problem, how is ROC AUC commonly extended?

Accepted Answer

One-vs-Rest or One-vs-One, then average the AUCs

Answer

Only the largest class is used

Answer

Multi-class AUC is impossible

Answer

By summing all confusion matrices into one number

Question 26

Two ROC curves cross each other. What does this imply about comparing the models by AUC alone?

Accepted Answer

AUC alone can hide region-specific trade-offs

Answer

The model with lower AUC is always better

Answer

Crossing curves means both AUCs are exactly 0.5

Answer

AUC comparison is always sufficient

Question 27

A classifier outputs the same probability for every instance. What will its ROC curve and AUC look like?

Accepted Answer

Diagonal line, AUC = 0.5

Answer

Top-left corner, AUC = 1.0

Answer

A vertical line, AUC = 0

Answer

Bottom-right curve, AUC = 0.25

Question 28

Why does each point on an ROC curve correspond to a confusion matrix?

Accepted Answer

Each threshold yields a distinct confusion matrix

Answer

Points are chosen randomly

Answer

Only one confusion matrix exists per model

Answer

Confusion matrices are unrelated to ROC

Question 29

What is the connection between AUC and the Gini coefficient in classification?

Accepted Answer

Gini = 2 * AUC - 1

Answer

Gini = AUC / 2

Answer

Gini = 1 - AUC

Answer

Gini = AUC squared

Question 30

When plotting an ROC curve in Python, why do we usually pass predicted probabilities rather than hard 0/1 labels?

Accepted Answer

Probabilities let the threshold sweep produce a full curve

Answer

Hard labels give a smoother curve

Answer

Probabilities are required by accuracy_score

Answer

Hard labels increase the AUC

Test yourself