DSE Supervised Learning: Classification 2 — Questions and Answers
Question 1: Which metric is most appropriate when false negatives are far more costly than false positives, such as in cancer screening?
- Precision
- Recall (Correct answer)
- Specificity
- Accuracy
Correct answer: Recall
Recall (sensitivity) measures how many actual positives are correctly identified, minimizing false negatives.
Question 2: In a Random Forest classifier, how does increasing the number of trees typically affect variance?
- Increases variance
- Decreases variance (Correct answer)
- Has no effect on variance
- First decreases then increases variance
Correct answer: Decreases variance
More trees in a Random Forest reduce variance by averaging more independent predictions, though with diminishing returns.
Question 3: What is the primary difference between hard and soft voting in ensemble classifiers?
- Hard voting uses probabilities; soft voting uses majority class
- Soft voting averages predicted probabilities; hard voting uses majority class labels (Correct answer)
- Hard voting is only for binary classification; soft voting works for multi-class
- Soft voting requires fewer estimators than hard voting
Correct answer: Soft voting averages predicted probabilities; hard voting uses majority class labels
Soft voting averages the class probabilities from each classifier, while hard voting picks the class predicted most frequently.
Question 4: Which kernel function in an SVM maps data into infinite-dimensional space?
- Linear kernel
- Polynomial kernel
- RBF (Gaussian) kernel (Correct answer)
- Sigmoid kernel
Correct answer: RBF (Gaussian) kernel
The RBF kernel implicitly maps data into an infinite-dimensional feature space via the Gaussian function.
Question 5: A logistic regression model outputs a probability of 0.45 for class 1. With a default threshold of 0.5, what class is predicted?
- Class 1
- Class 0 (Correct answer)
- The model abstains
- Depends on the training data
Correct answer: Class 0
Since 0.45 is below the 0.5 threshold, the model predicts class 0.
Question 6: What does the term 'class imbalance' refer to in classification?
- The model predicts one class more slowly than another
- One class has significantly more samples than other classes in the training data (Correct answer)
- The decision boundary is not centered between classes
- Feature scales differ between classes
Correct answer: One class has significantly more samples than other classes in the training data
Class imbalance occurs when the distribution of target classes in the dataset is skewed, with one class having far more samples.
Question 7: Which technique specifically addresses class imbalance by generating synthetic samples for the minority class?
- Bagging
- SMOTE (Correct answer)
- L2 regularization
- Feature scaling
Correct answer: SMOTE
SMOTE (Synthetic Minority Over-sampling Technique) creates new synthetic samples by interpolating between existing minority class examples.
DSE Supervised Learning: Classification 2
Which metric is most appropriate when false negatives are far more costly than false positives, such as in cancer screening?