ROC and AUC curves in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding ROC Curve Basics

What does the ROC curve represent in a binary classification model?

A The relationship between precision and recall for different classes

B The trade-off between true positive rate and false positive rate at different thresholds

C The distribution of predicted probabilities for positive class only

D The accuracy of the model on the training dataset

💡 Hint

Think about what happens when you change the decision threshold in classification.

The ROC curve plots the true positive rate (sensitivity) against the false positive rate (1 - specificity) at various threshold settings, showing the trade-off between detecting positives and avoiding false alarms.

❓ Predict Output

intermediate

Output of TensorFlow AUC Metric Calculation

What is the output of the following TensorFlow code snippet?

TensorFlow

import tensorflow as tf

labels = [0, 0, 1, 1]
predictions = [0.1, 0.4, 0.35, 0.8]
auc_metric = tf.keras.metrics.AUC()
auc_metric.update_state(labels, predictions)
result = auc_metric.result().numpy()
print(round(result, 2))

A 0.95

B 0.85

C 0.65

D 0.75

💡 Hint

Calculate the area under the ROC curve for the given predictions and labels.

The AUC value here is 0.75, representing the probability that a randomly chosen positive instance ranks higher than a randomly chosen negative one.

❓ Model Choice

advanced

Choosing the Best Model Based on ROC AUC

You trained three binary classifiers and obtained these ROC AUC scores on the validation set: Model A: 0.82, Model B: 0.91, Model C: 0.88. Which model should you select if you want the best overall ability to distinguish classes?

A Model B

B All models perform equally well

C Model C

D Model A

💡 Hint

Higher AUC means better class separation ability.

Model B has the highest ROC AUC score (0.91), indicating it best separates positive and negative classes.

❓ Hyperparameter

advanced

Effect of Threshold on ROC Curve

How does changing the classification threshold affect the ROC curve?

A It moves the point along the ROC curve, changing true positive and false positive rates

B It changes the shape of the ROC curve itself

C It only affects precision but not the ROC curve

D It has no effect on ROC curve metrics

💡 Hint

Think about what happens when you decide to be more or less strict in classifying positives.

Adjusting the threshold moves the operating point along the ROC curve, altering true positive and false positive rates but not the curve shape.

🔧 Debug

expert

Identifying the Bug in AUC Calculation Code

What error does the following TensorFlow code produce?

import tensorflow as tf

labels = [0, 1, 0, 1]
predictions = [0.2, 0.8, 0.4]
auc_metric = tf.keras.metrics.AUC()
auc_metric.update_state(labels, predictions)
print(auc_metric.result().numpy())

A No error, outputs a float value

B TypeError because predictions are not tensors

C ValueError due to mismatched lengths of labels and predictions

D AttributeError because AUC metric is not imported correctly

💡 Hint

Check if labels and predictions have the same number of elements.

The code raises a ValueError because labels and predictions lists have different lengths (4 vs 3), which is invalid for metric calculation.