1.

Introduction to Classification

1. What is classification in machine learning?

o a) Predicting categorical labels ✅

o b) Predicting continuous values

o c) Clustering data

o d) Reducing dimensionality

2. Which of the following is NOT a classification algorithm?

o a) K-Means ✅

o b) K-Nearest Neighbors (KNN)

o c) Decision Tree

o d) Naïve Bayes

3. What is the key difference between classification and regression?

o a) Classification predicts discrete labels, while regression predicts continuous

o b) Classification works only with text data

o c) Regression is a type of unsupervised learning

o d) Classification is only used in deep learning

4. What is the purpose of a confusion matrix in classification?

o a) To evaluate the performance of a classifier ✅

o b) To reduce the number of features

o c) To perform regression analysis

o d) To generate synthetic data

5. What is precision in classification?

o a) The ratio of true positives to the sum of true positives and false positives ✅

o b) The ratio of true negatives to total instances

o c) The fraction of correctly predicted positive instances

2. K-Nearest Neighbors (KNN)

6. K-Nearest Neighbors (KNN) is classified as a:

o a) Instance-based learning algorithm ✅

o b) Parametric algorithm

o c) Deep learning model

o d) Reinforcement learning technique

7. How does KNN classify a new data point?

o a) By finding the most common class among the k nearest neighbors ✅

o b) By calculating decision boundaries

o c) By training a neural network

o d) By creating decision trees

8. What happens when the value of K in KNN is too large?

o a) The model becomes less flexible and may underfit ✅

o b) The model overfits the data

o c) The model ignores new data points

o d) The model stops working

9. What is the best way to choose the optimal K value in KNN?

o a) Using cross-validation ✅

o b) Random selection

o c) Using a fixed value of K=10

o d) Increasing K indefinitely

10. What type of distance metric is commonly used in KNN?

 a) Euclidean distance ✅

 b) Cosine similarity
  c) Manhattan distance

 d) Jaccard distance

3. Decision Tree Classifier

11. Decision trees work by:

 a) Splitting data based on feature values ✅

 b) Assigning weights to features

 c) Calculating probability distributions

 d) Using backpropagation

12. What is a leaf node in a decision tree?

 a) A node that represents a class label ✅

 b) The starting node

 c) A node that splits into two

 d) A randomly selected node

13. What is overfitting in decision trees?

 a) When the model is too complex and learns noise ✅

 b) When the model is too simple

 c) When the model performs well on new data

 d) When the tree has too few nodes

14. What is entropy in a decision tree?

 a) A measure of uncertainty in the dataset ✅

 b) The number of branches in a tree

 c) The height of the tree

 d) The count of decision nodes

15. Which algorithm is commonly used to construct decision trees?

 a) ID3 ✅
  b) K-Means

 c) Gradient Boosting

 d) PCA

4. Naïve Bayes Classifier

16. What assumption does Naïve Bayes make?

 a) Features are independent ✅

 b) Features are dependent

 c) Data is linearly separable

 d) The dataset is balanced

17. Naïve Bayes is based on which theorem?

 a) Bayes' theorem ✅

 b) Pythagorean theorem

 c) Markov property

 d) Euclidean distance

18. What is the advantage of Naïve Bayes?

 a) Works well with small datasets ✅

 b) Requires a lot of training data

 c) Only works with numerical data

 d) Ignores missing values

19. In text classification, Naïve Bayes is commonly used because:

 a) It handles high-dimensional data well ✅

 b) It does not require labeled data

 c) It works only with binary data

 d) It needs a deep learning model

20. What is Laplace smoothing in Naïve Bayes?

 b) A way to improve accuracy

 c) A method to reduce variance

 d) A feature selection technique

5. Support Vector Machine (SVM)

21. What is the key idea behind SVM?

 a) Finding the best hyperplane that separates classes ✅

 b) Calculating mean values

 c) Creating decision trees

 d) Generating random clusters

22. What is a support vector in SVM?

 a) A data point closest to the decision boundary ✅

 b) The largest dataset sample

 c) A random point in the dataset

 d) The center of a cluster

23. What kernel function is commonly used in SVM for non-linear classification?

 a) Radial Basis Function (RBF) ✅

 b) Linear kernel

 c) Manhattan kernel

 d) Euclidean kernel

24. What does a soft-margin SVM do?

 a) Allows some misclassification to avoid overfitting ✅

 b) Only works with linearly separable data

 c) Uses deep learning

 d) Reduces dataset size

 a) C (Regularization parameter) ✅

 b) K (Neighbors)

 c) Entropy

 d) Tree depth

6. Regression (Linear & Polynomial)

26. What is linear regression used for?

 a) Predicting continuous values ✅

 b) Predicting categorical values

 c) Reducing dimensionality

 d) Clustering data

27. What is the equation for simple linear regression?

 a) y = mx + b ✅

 b) y = a + bx²

 c) y = x / 2

 d) y = log(x)

28. What is the difference between polynomial and linear regression?

 a) Polynomial regression fits a curve, linear regression fits a straight line ✅

 b) Polynomial regression is faster

 c) Linear regression requires more data

 d) They are the same

29. What is R-squared in regression?

 a) A measure of how well the model fits the data ✅

 b) The sum of squared errors

 c) The correlation coefficient

30. What is multicollinearity in regression?

 a) When predictor variables are highly correlated ✅

 b) When the dataset is too large

 c) When data has missing values

 d) When classification is used instead of regression

31. Which evaluation metric is most appropriate for imbalanced classification problems?

 a) F1-score ✅

 b) Accuracy

 c) Mean Squared Error

 d) R-squared

32. What is the ROC curve used for?

 a) To visualize the performance of a classification model ✅

 b) To select the best K value in KNN

 c) To measure regression accuracy

 d) To determine the optimal number of clusters

33. In binary classification, what does a high recall value indicate?

 a) The model correctly identifies most positive instances ✅

 b) The model is very specific

 c) The model always predicts "negative"

 d) The model has low accuracy

34. What is a common way to handle class imbalance in a dataset?

 a) Oversampling the minority class ✅

 b) Removing the majority class

 c) Reducing the dataset size

 d) Ignoring the imbalance

 a) Naïve Bayes ✅

 b) Decision Tree

 c) KNN

 d) SVM

32. K-Nearest Neighbors (KNN)

36. What happens when K=1 in KNN?

 a) The model becomes highly sensitive to noise (overfitting) ✅

 b) The model becomes more generalized

 c) The model ignores distance

 d) The model cannot classify data

37. What is the time complexity of KNN for prediction?

 a) O(n) ✅

 b) O(log n)

 c) O(1)

 d) O(n²)

38. Which distance metric is NOT commonly used in KNN?

 a) Jaccard similarity ✅

 b) Euclidean distance

 c) Manhattan distance

 d) Minkowski distance

39. How can we speed up KNN classification?

 a) Using KD-trees or Ball-trees ✅

 b) Reducing the training set

 c) Increasing the K value indefinitely

40. Which library is commonly used for implementing KNN in Python?

 a) Scikit-learn ✅

 b) TensorFlow

 c) PyTorch

 d) Pandas

33. Decision Trees

41. Which criterion is commonly used to split nodes in Decision Trees?

 a) Gini impurity or entropy ✅

 b) R-squared

 c) Euclidean distance

 d) Cosine similarity

42. What is pruning in decision trees?

 a) Removing unnecessary branches to prevent overfitting ✅

 b) Adding more branches to improve accuracy

 c) Reducing the dataset size

 d) Changing decision thresholds

43. How does a random forest improve over a single decision tree?

 a) By averaging multiple trees to reduce overfitting ✅

 b) By using only a single feature for classification

 c) By increasing depth indefinitely

 d) By selecting only one training sample

44. Which of the following is a major limitation of Decision Trees?

 a) They tend to overfit on small datasets ✅

 b) They cannot handle categorical data

 d) They require deep learning

45. What is the main advantage of using Decision Trees?

 a) They are easy to interpret and understand ✅

 b) They always achieve 100% accuracy

 c) They require less data preprocessing

 d) They work only with large datasets

34. Naïve Bayes

46. What is the main assumption of Naïve Bayes?

 a) Features are conditionally independent given the class ✅

 b) Features are dependent

 c) Classes are equally distributed

 d) All data points are equally important

47. Why is Naïve Bayes commonly used for spam detection?

 a) It performs well on high-dimensional text data ✅

 b) It ignores probability distributions

 c) It requires deep learning

 d) It only works on numerical data

48. Which type of Naïve Bayes classifier is best suited for text classification?

 a) Multinomial Naïve Bayes ✅

 b) Gaussian Naïve Bayes

 c) Bernoulli Naïve Bayes

 d) Logistic Regression

49. When does Naïve Bayes perform poorly?

 a) When features are highly correlated ✅

 c) When the dataset is balanced

 d) When using numerical features

50. How is probability calculated in Naïve Bayes?

 a) Using Bayes’ theorem ✅

 b) Using K-means clustering

 c) Using gradient descent

 d) Using entropy reduction

51. Which type of learning is classification?

 a) Supervised learning ✅

 b) Unsupervised learning

 c) Reinforcement learning

 d) Semi-supervised learning

52. What is stratified sampling in classification?

 a) Ensuring each class is proportionally represented in training and test sets ✅

 b) Randomly selecting data points

 c) Increasing dataset size

 d) Using a single training sample

53. What happens if a classification model has high bias?

 a) It underfits the data ✅

 b) It overfits the data

 c) It has high accuracy

 d) It performs well on test data

54. What is a One-vs-All (OvA) approach in multi-class classification?

 a) Training a separate binary classifier for each class ✅

 b) Combining all classes into a single label

 d) Ignoring minority classes

55. What is the impact of an imbalanced dataset in classification?

 a) The model may favor the majority class ✅

 b) The model performs better

 c) It reduces overfitting

 d) It improves generalization

52. K-Nearest Neighbors (KNN)

56. How is the decision boundary in KNN defined?

 a) By the nearest neighbors' class labels ✅

 b) By a fixed linear function

 c) By entropy

 d) By a hyperplane

57. Which of the following is a drawback of KNN?

 a) It is slow for large datasets ✅

 b) It requires a lot of training

 c) It cannot handle numerical data

 d) It only works for regression

58. What effect does increasing K have on KNN?

 a) It smooths decision boundaries ✅

 b) It increases overfitting

 c) It ignores test data

 d) It reduces bias

59. Which technique can help reduce computational cost in KNN?

 a) KD-trees ✅
  b) Increasing dataset size

 c) Ignoring distances

 d) Removing test data

60. What is the best value of K in KNN?

 a) Depends on the dataset and should be chosen using cross-validation ✅

 b) Always 5

 c) Always 1

 d) Any odd number

53. Decision Trees

61. What is the stopping criterion in decision trees?

 a) When further splitting does not improve purity ✅

 b) When all features are used

 c) When entropy is maximized

 d) When dataset size is reduced

62. How does a decision tree handle missing values?

 a) By estimating based on other feature values ✅

 b) By ignoring missing values

 c) By stopping training

 d) By removing all data points with missing values

63. What is information gain in decision trees?

 a) The reduction in entropy after a split ✅

 b) The depth of the tree

 c) The size of the dataset

 d) The learning rate

64. Which method can help prevent overfitting in decision trees?

 b) Increasing tree depth

 c) Using all features

 d) Using linear regression

65. What is a major limitation of decision trees?

 a) They are prone to overfitting on small datasets ✅

 b) They cannot handle categorical data

 c) They are too simple

 d) They cannot be interpreted

54. Naïve Bayes

66. What is the main limitation of Naïve Bayes?

 a) It assumes feature independence, which is not always true ✅

 b) It cannot handle text data

 c) It requires large datasets

 d) It only works with numerical data

67. Which Naïve Bayes variant is best for binary classification?

 a) Bernoulli Naïve Bayes ✅

 b) Multinomial Naïve Bayes

 c) Gaussian Naïve Bayes

 d) Logistic Regression

68. Why is Naïve Bayes fast compared to other classifiers?

 a) It does not require complex training ✅

 b) It uses deep learning

 c) It ignores feature distributions

 d) It always predicts the majority class

 a) Using Gaussian (Normal) distribution assumption ✅

 b) By converting it to categorical data

 c) By ignoring it

 d) By using decision trees

70. What happens if Naïve Bayes encounters a zero probability for a category?

 a) It applies Laplace smoothing ✅

 b) It ignores the category

 c) It stops predicting

 d) It increases the dataset size