DSE Supervised Learning: Classification 5 — Questions and Answers
Question 1: What is 'concept drift' in the context of a deployed classification model?
- The model's weights change due to continued training on new data
- The statistical properties of the target variable change over time, degrading model performance (Correct answer)
- The feature engineering pipeline introduces new variables
- The model begins to memorize training data after deployment
Correct answer: The statistical properties of the target variable change over time, degrading model performance
Concept drift occurs when the relationship between input features and the target class shifts over time in production, causing model accuracy to degrade.
Question 2: In a confusion matrix for binary classification, what does a False Positive (FP) represent?
- A negative instance correctly classified as negative
- A negative instance incorrectly classified as positive (Correct answer)
- A positive instance correctly classified as positive
- A positive instance incorrectly classified as negative
Correct answer: A negative instance incorrectly classified as positive
A False Positive occurs when the model predicts the positive class but the true label is negative.
Question 3: What is the purpose of calibrating a classifier's predicted probabilities?
- To improve the model's accuracy on the training set
- To ensure the predicted probabilities accurately reflect true likelihoods (Correct answer)
- To reduce the number of features used by the model
- To make the decision boundary linear
Correct answer: To ensure the predicted probabilities accurately reflect true likelihoods
Calibration aligns predicted probabilities with actual observed frequencies, so a model predicting 0.8 is correct about 80% of the time.
Question 4: Which ensemble method trains base classifiers sequentially, with each model focusing more on previously misclassified examples?
- Bagging
- Boosting (Correct answer)
- Stacking
- Voting
Correct answer: Boosting
Boosting trains classifiers sequentially, reweighting training samples so subsequent models focus on the errors of prior ones.
Question 5: What is the 'curse of dimensionality' and how does it affect k-NN classification?
- High dimensionality speeds up k-NN by reducing search space
- As dimensions increase, distances between points become similar, making nearest neighbors less meaningful (Correct answer)
- k-NN requires exponentially more classes as dimensions grow
- High dimensionality causes k-NN to overfit via the kernel trick
Correct answer: As dimensions increase, distances between points become similar, making nearest neighbors less meaningful
In high-dimensional spaces, Euclidean distances concentrate, making it hard to distinguish 'near' from 'far' neighbors, which undermines k-NN's core assumption.
Question 6: Which of the following activation functions is used in the output layer of a neural network for binary classification?
- ReLU
- Sigmoid (Correct answer)
- Softmax
- Tanh
Correct answer: Sigmoid
The sigmoid function maps the output to a value between 0 and 1, which can be interpreted as the probability of the positive class.
Question 7: A model trained on historical loan data performs well on training data but poorly on new applicants. Which problem does this most likely indicate?
- Underfitting due to high bias
- Overfitting due to high variance (Correct answer)
- Class imbalance in the test set only
- An incorrect choice of evaluation metric
Correct answer: Overfitting due to high variance
Strong training performance but poor generalization to new data is the classic symptom of overfitting, where the model has learned noise or patterns specific to the training set.
DSE Supervised Learning: Classification 5
What is 'concept drift' in the context of a deployed classification model?