Back to Subject

Unit 3: Mining frequent patterns - Practice Quiz

BTY587 — Data Analysis And Simulations 60 Questions
0 Correct 0 Wrong 60 Left
0/60

1 In data mining, an association between items refers to:

associations and correlations Easy
A. A way to encrypt transaction records
B. A method to sort items alphabetically
C. A technique to compress large datasets
D. A relationship where items tend to occur together in transactions

2 A correlation measures:

associations and correlations Easy
A. The speed of a mining algorithm
B. The statistical relationship or dependency between two variables
C. The total number of transactions in a database
D. The physical storage size of a dataset

3 Which of the following best describes the difference between association and correlation?

associations and correlations Easy
A. Association shows co-occurrence, while correlation measures strength and direction of dependency
B. They are exactly the same concept
C. Correlation is used only for encryption
D. Association applies only to numbers, correlation only to text

4 An association rule is typically written in the form:

Introduction to Association Rule Mining Easy
A.
B.
C.
D.

5 The support of an itemset is defined as:

Introduction to Association Rule Mining Easy
A. The order in which items are scanned
B. The average price of items
C. The fraction of transactions that contain the itemset
D. The number of columns in the dataset

6 The confidence of the rule is calculated as:

Introduction to Association Rule Mining Easy
A.
B.
C.
D.

7 Which classic algorithm is widely used to mine frequent itemsets for association rules?

Introduction to Association Rule Mining Easy
A. QuickSort
B. K-Means
C. Apriori
D. Dijkstra

8 A frequent itemset is one whose support is:

Introduction to Association Rule Mining Easy
A. Always equal to zero
B. Equal to the number of attributes
C. Less than the average transaction size
D. Greater than or equal to a minimum support threshold

9 Which measure is commonly used to evaluate correlation between items in association analysis?

Mining Correlation Patterns Easy
A. Lift
B. Latency
C. Radius
D. Depth

10 A lift value equal to indicates that items and are:

Mining Correlation Patterns Easy
A. Independent of each other
B. Impossible to occur together
C. Perfectly positively correlated
D. Perfectly negatively correlated

11 A lift value greater than suggests that two items are:

Mining Correlation Patterns Easy
A. Completely unrelated
B. Positively correlated
C. Negatively correlated
D. Never purchased together

12 The main reason for mining correlation patterns rather than only association rules is to:

Mining Correlation Patterns Easy
A. Encrypt the itemsets for security
B. Reduce the total number of transactions stored
C. Speed up disk read operations
D. Identify whether an association is statistically meaningful and not just frequent

13 Classification is a type of learning that is:

Introduction to Classification Easy
A. Used only for encryption
B. Unsupervised, using no labels
C. Based only on clustering
D. Supervised, using labeled training data

14 The output of a classification model is:

Introduction to Classification Easy
A. A continuous numeric value
B. A sorted list of items
C. A discrete class label
D. A compressed file

15 Which of the following is an example of a classification task?

Introduction to Classification Easy
A. Sorting numbers in ascending order
B. Grouping customers without any labels
C. Predicting tomorrow's exact temperature in degrees
D. Predicting whether an email is spam or not spam

16 Frequent pattern-based classification uses which of the following as features?

Frequent Pattern-Based Classification Methods Easy
A. Randomly generated numbers
B. Frequent patterns discovered from the data
C. Only the class labels themselves
D. Encrypted transaction IDs

17 An associative classifier builds classification rules using techniques adapted from methods that were originally designed to mine frequent itemsets and generate association rules from transaction databases.

Frequent Pattern-Based Classification Methods Easy
A. Image compression
B. Association rule mining
C. Linear regression
D. Sorting algorithms

18 Precision is defined as:

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Easy
A.
B.
C.
D.

19 Recall is defined as:

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Easy
A.
B.
C.
D.

20 The F1-Score is the harmonic mean of:

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Easy
A. Precision and Recall
B. True Positives and True Negatives
C. Lift and Confidence
D. Accuracy and Support

21 A transaction database has 1000 transactions. The itemset appears in 150 transactions. What is the support of this itemset?

Introduction to Association Rule Mining Medium
A.
B.
C.
D.

22 For the rule , support() and support() . What is the confidence of the rule?

Introduction to Association Rule Mining Medium
A.
B.
C.
D.

23 A rule has a lift value of . What does this indicate?

associations and correlations Medium
A. and are independent
B. and are positively correlated
C. and are negatively correlated
D. The rule has 100% confidence

24 Given support() , support() , and support() , what is the lift of ?

associations and correlations Medium
A.
B.
C.
D.

25 The Apriori algorithm relies on which key property to reduce the search space?

Introduction to Association Rule Mining Medium
A. If an itemset is frequent, all its subsets are frequent
B. If an itemset is infrequent, all its supersets are frequent
C. Confidence always increases with itemset size
D. All itemsets have equal support

26 A rule with high confidence but lift less than suggests which relationship between the antecedent and consequent?

Mining Correlation Patterns Medium
A. Positive correlation
B. Strong causation
C. Perfect independence
D. Negative correlation

27 Which correlation measure is preferred over lift because it is null-invariant (not affected by transactions containing neither item)?

Mining Correlation Patterns Medium
A. Cosine measure
B. Support
C. Confidence
D. Lift

28 In a correlation test for two items, a value significantly greater than the expected threshold indicates that the items are:

Mining Correlation Patterns Medium
A. Independent
B. Correlated (dependent)
C. Mutually exclusive only
D. Equally frequent

29 Which of the following best distinguishes classification from clustering?

Introduction to Classification Medium
A. Clustering predicts continuous values
B. Classification is unsupervised; clustering is supervised
C. Both require predefined class labels
D. Classification uses labeled training data; clustering does not

30 In a two-step classification process, what is the purpose of the first (learning) step?

Introduction to Classification Medium
A. Build a model from the training set
B. Measure the model's accuracy
C. Apply the model to unseen data
D. Remove noisy attributes only

31 Why is a separate test set used instead of evaluating a classifier on its training data?

Introduction to Classification Medium
A. To speed up model building
B. To guarantee zero classification error
C. To increase the training accuracy
D. To obtain an unbiased estimate of generalization performance

32 Associative classification builds a classifier primarily from which type of rules?

Frequent Pattern-Based Classification Methods Medium
A. Random decision splits
B. Class association rules of the form
C. Regression equations
D. Unlabeled cluster centroids

33 In frequent pattern-based classification, why are discriminative frequent patterns preferred over all frequent patterns?

Frequent Pattern-Based Classification Methods Medium
A. They reduce confidence to zero
B. They always have the highest support
C. They provide higher class-discriminating power as features
D. They eliminate the need for training data

34 The CBA (Classification Based on Associations) algorithm ranks its rules primarily by which criteria?

Frequent Pattern-Based Classification Methods Medium
A. Support, then lift
B. Confidence, then support
C. Rule length, then confidence
D. Alphabetical order of items

35 A classifier gives TP , FP , FN , TN . What is the precision?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Medium
A.
B.
C.
D.

36 Using TP , FP , FN , TN , what is the recall?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Medium
A.
B.
C.
D.

37 If a model has precision and recall , what is its F1-score?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Medium
A.
B.
C.
D.

38 Why is the F1-score often preferred over accuracy for imbalanced datasets?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Medium
A. It ignores false negatives entirely
B. It balances precision and recall, ignoring true negatives
C. It counts true negatives most heavily
D. It always equals accuracy

39 A spam filter that marks almost all emails as spam will most likely have:

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Medium
A. High precision but low recall
B. High precision and high recall
C. High recall but low precision
D. Low recall and low precision

40 A rule can be 'strong' (high support and confidence) yet misleading. Which measure best exposes this by comparing observed to expected co-occurrence?

associations and correlations Medium
A. Confidence
B. Coverage
C. Support
D. Lift

41 A transaction database has 10,000 transactions. Itemset appears in 500 transactions, item appears in 2,000, and item appears in 1,000. For the rule , what are the support and confidence respectively?

Introduction to Association Rule Mining Hard
A. Support , Confidence
B. Support , Confidence
C. Support , Confidence
D. Support , Confidence

42 For the rule with , , and , what is the lift, and what does it imply?

associations and correlations Hard
A. Lift ; and are negatively correlated
B. Lift ; and are independent
C. Lift ; and are positively correlated
D. Lift ; and are negatively correlated

43 Consider a dataset in which 90% of all transactions contain milk. A rule has confidence 0.85. What is the most accurate interpretation?

Introduction to Association Rule Mining Hard
A. The rule guarantees that 85% of milk buyers also buy bread
B. The rule is misleading; buying bread actually makes buying milk slightly less likely than baseline
C. The rule is strong because confidence exceeds 0.8
D. The rule shows bread strongly implies milk since 0.85 is a high confidence

44 The lift measure is not null-invariant. Which of the following measures is null-invariant, making it more suitable for large sparse datasets with many null transactions?

Mining Correlation Patterns Hard
A. Support
B. Cosine measure
C. Lift
D. (chi-square)

45 For two items with , , and , compute the cosine measure.

Mining Correlation Patterns Hard
A.
B.
C.
D.

46 A binary classifier produces the confusion matrix: TP = 40, FP = 10, FN = 20, TN = 30. What is the F1-score?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Hard
A.
B.
C.
D.

47 In a highly imbalanced dataset where the positive class is only 2% of the data, a model predicts every instance as negative. Which statement is correct?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Hard
A. Accuracy is 98% but recall for the positive class is 0
B. Accuracy is 98% and F1-score for the positive class is also high
C. Both accuracy and F1-score for the positive class are 98%
D. Accuracy is 2% and precision is undefined but recall is 100%

48 The score generalizes F1. If a fraud-detection team wants to weight recall twice as heavily as precision, which should they use?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Hard
A.
B.
C.
D.

49 Associative classification methods such as CBA build classifiers from class association rules (CARs). What is the key difference between a CAR and a general association rule?

Frequent Pattern-Based Classification Methods Hard
A. A CAR uses only single-item antecedents
B. A CAR requires confidence to always equal 1
C. A CAR has no minimum support threshold
D. The consequent of a CAR is restricted to a class label

50 Why can discriminative frequent patterns improve classification accuracy over using single features, even though the number of frequent patterns can be huge?

Frequent Pattern-Based Classification Methods Hard
A. Frequent patterns eliminate the need for any feature selection at all
B. Frequent patterns always have higher support than individual features and are therefore more reliable
C. Frequent patterns capture feature combinations that carry more discriminative information than individual low-order features
D. Frequent patterns reduce the dataset dimensionality to exactly the number of classes present

51 In classification, why is a purely training-error-minimizing model prone to overfitting?

Introduction to Classification Hard
A. It always underestimates the number of classes present
B. It cannot compute a decision boundary for linearly separable data
C. It may model noise in the training data, reducing generalization to unseen data
D. It requires that training and test distributions be identical

52 The Apriori algorithm's efficiency relies on the downward closure property. Which statement correctly expresses this property?

Introduction to Association Rule Mining Hard
A. All supersets of a frequent itemset are also frequent
B. If an itemset is infrequent then all its subsets are infrequent
C. All non-empty subsets of a frequent itemset are also frequent
D. The support of an itemset equals the sum of its subsets' supports

53 Two items have a contingency table: both present = 400, A only = 100, B only = 100, neither = 400 (total = 1000). Testing independence with , what is the expected count for the 'both present' cell?

Mining Correlation Patterns Hard
A.
B.
C.
D.

54 The Kulczynski measure combined with imbalance ratio (IR) is often used to judge interestingness. When Kulczynski and IR is high, what does this typically indicate?

associations and correlations Hard
A. A strongly positively correlated and balanced pattern
B. A neutral/uncorrelated pattern that is also strongly imbalanced between the two items
C. A strongly negatively correlated pattern with balanced item frequencies
D. A guaranteed spurious rule that must be discarded regardless of IR

55 For a multiclass problem, macro-averaged F1 and micro-averaged F1 can differ substantially. When is micro-F1 preferable?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Hard
A. When every class should contribute equally regardless of size
B. When you only care about the rarest minority class
C. When the dataset has exactly balanced class frequencies
D. When you want performance dominated by the more frequent classes

56 In the CMAR (Classification based on Multiple Association Rules) approach, when multiple matching rules predict different classes for a test instance, how is the conflict typically resolved?

Frequent Pattern-Based Classification Methods Hard
A. By always choosing the single rule with the highest support
B. By taking a simple unweighted majority vote of all matching rules
C. By grouping matching rules per class and using a weighted measure to select the strongest group
D. By choosing the rule that was generated earliest during mining

57 A closed frequent itemset is defined such that no immediate superset has the same support. Why are closed itemsets preferred over the full set of frequent itemsets?

Introduction to Association Rule Mining Hard
A. They guarantee all association rules will have confidence 1
B. They provide a lossless, compact representation from which all frequent itemsets and their supports can be recovered
C. They are always fewer than the number of maximal itemsets
D. They require less memory because they discard support counts entirely

58 A lazy learner like k-NN differs fundamentally from an eager learner like a decision tree in that it:

classification Hard
A. Builds a global model during training and discards the training data
B. Defers all generalization until a query arrives, storing the training data instead of building a model
C. Cannot handle numeric attributes without discretization
D. Always achieves lower test error than eager learners

59 Model X has precision 0.9, recall 0.6; Model Y has precision 0.7, recall 0.8. Comparing by F1-score, which model wins and by roughly how much?

Pattern Evaluation in Classification (Evaluation Metrics: Precision, Recall, F1-Score) Hard
A. Model X wins ( vs )
B. They tie at
C. Model Y wins ( vs )
D. Model X wins ( vs )

60 The classic 'coffee-tea' example shows that the rule can have high confidence yet lift . What is the primary lesson from this paradox?

associations and correlations Hard
A. Support must be increased to fix the contradiction between confidence and lift
B. Lift should never be used because it contradicts confidence
C. High confidence does not imply positive correlation; correlation measures like lift are needed to detect genuine dependence
D. Confidence and lift always move in the same direction
Previous MCQ
Unit 2
Next MCQ
Unit 4