CDSP 5 Quiz

1. What is binary classification?

A: Categorizing data into multiple classes
B: Categorizing data examples as either a 1 or a 0
C: Classifying images into different categories
D: Classifying data without using labels

2. What is the purpose of logistic regression?

A: To predict continuous values
B: To classify data based on probabilities
C: To reduce the number of features
D: To cluster data into groups

3. What is a decision boundary in logistic regression?

A: A point where the model stops learning
B: A threshold used to determine class membership
C: A method to reduce model overfitting
D: A way to visualize multi-class classification

4. What is multi-label classification?

A: Classifying data into more than two mutually exclusive classes
B: Classifying data with multiple labels assigned to each instance
C: A type of clustering algorithm
D: Classifying data without any supervision

5. What function is used in multinomial logistic regression to determine class probabilities?

A: Sigmoid function
B: Softmax function
C: ReLU function
D: Decision function

6. What is k-nearest neighbor (k-NN) used for in classification?

A: Determining the most relevant features
B: Predicting class labels based on the closest data points
C: Identifying outliers in the dataset
D: Reducing the dimensionality of the data

7. What happens when the value of k is increased in k-NN?

A: The model becomes more sensitive to outliers
B: The boundaries between classes become less distinct
C: The model overfits the training data
D: The model becomes less computationally expensive

8. What is the purpose of support-vector machines (SVMs) in classification?

A: To minimize the number of classes
B: To find a hyperplane that maximizes the margin between classes
C: To cluster data into separate groups
D: To reduce the size of the dataset

9. What is the key feature of soft-margin classification in SVMs?

A: It allows some misclassified points within the margin
B: It creates non-linear decision boundaries
C: It only works with large datasets
D: It requires data to be linearly separable

10. What is the primary assumption of Naïve Bayes classifiers?

A: Features are dependent on each other
B: Features are conditionally independent
C: All features contribute equally to the prediction
D: The model uses unsupervised learning

11. What does the Gini index measure in decision trees?

A: The complexity of the model
B: The accuracy of the classification
C: The impurity of a node
D: The size of the dataset

12. What is the purpose of pruning in decision trees?

A: To increase the number of branches
B: To reduce the size of the tree and avoid overfitting
C: To improve the depth of the tree
D: To increase the number of features used

13. What is the primary advantage of random forests over decision trees?

A: They provide more accurate predictions
B: They require less computational power
C: They are less prone to overfitting
D: They use fewer features

14. What is the goal of gradient boosting?

A: To increase the size of the decision tree
B: To build weak learners that improve the previous model's errors
C: To reduce the dimensionality of the data
D: To select the best features for the model

15. What is hyperparameter optimization in machine learning?

A: Automatically adjusting model parameters during training
B: Manually tuning the learning rate
C: Repeatedly altering hyperparameters to find the best model performance
D: Changing the model architecture

16. What does the F₁ score represent in model evaluation?

A: The tradeoff between precision and recall
B: The overall accuracy of the model
C: The probability of false negatives
D: The true positive rate

17. What does a confusion matrix display?

A: The true positives and false negatives
B: The overall accuracy of a model
C: The number of features used in training
D: The relationship between model complexity and accuracy

18. What is a key difference between k-NN and logistic regression?

A: k-NN is a supervised learning algorithm, while logistic regression is unsupervised
B: k-NN does not learn parameters during training, while logistic regression does
C: Logistic regression handles outliers better than k-NN
D: k-NN requires less computational power than logistic regression

19. What is the purpose of cross-validation in model training?

A: To increase the size of the training set
B: To evaluate model performance on unseen data
C: To generate more data for the model
D: To automatically select the best model

20. What is an ensemble learning method that trains multiple weak learners to form a strong learner?

A: Decision trees
B: k-means clustering
C: Gradient boosting
D: Principal component analysis