ADVANCED DATA MINING
COMPREHENSIVE QUESTIONS AND
ANSWERS. VERIFIED 2026/2027
1. In the context of data preprocessing, which of the following is an effect of the ‘Curse of
Dimensionality’?
A. Data points become more concentrated in the center of the space.
B. Computational efficiency increases due to the sparsity of the data.
C. The distance between any two points in high-dimensional space becomes nearly
equidistant.
D. Feature selection becomes redundant as all dimensions contribute equally to the
variance.
Answer: C
Conceptual Explanation: As dimensionality increases, the contrast between the distance
to the nearest and farthest neighbor decreases, making distance-based algorithms like KNN
less effective.
2. Which measure is most appropriate for assessing the similarity between two sparse binary
vectors (e.g., transaction data)?
A. Euclidean Distance
,B. Manhattan Distance
C. Pearson Correlation
D. Jaccard Coefficient
Answer: D
Conceptual Explanation: The Jaccard Coefficient is ideal for asymmetric binary attributes
because it ignores ‘0-0’ matches, which are often uninformative in sparse data.
3. What is the primary difference between Principal Component Analysis (PCA) and Linear
Discriminant Analysis (LDA)?
A. PCA is supervised while LDA is unsupervised.
B. LDA reduces dimensionality while PCA only transforms the feature space.
C. PCA seeks to maximize variance, whereas LDA seeks to maximize class separability.
D. PCA handles non-linear relationships whereas LDA is strictly linear.
Answer: C
Conceptual Explanation: PCA is an unsupervised technique that finds directions of
maximum variance. LDA is supervised and finds a projection that maximizes the ratio of
between-class variance to within-class variance.
4. In decision tree induction, what does a Gini Index of 0 indicate for a specific node?
A. The node is a leaf node and is perfectly pure.
B. The node contains an equal distribution of all classes.
, C. The node has reached the maximum allowed depth.
D. The split at this node is considered the worst possible split.
Answer: A
Conceptual Explanation: A Gini Index of 0 means all records in the node belong to a single
class, representing maximum purity.
5. Which of the following describes the ‘Anti-monotone’ property of the Support measure in
association rule mining?
A. Support increases as the number of items in the itemset increases.
B. If an itemset is infrequent, then all of its supersets must be frequent.
C. The confidence of a rule is always less than or equal to its support.
D. If an itemset is frequent, then all of its subsets must also be frequent.
Answer: D
Conceptual Explanation: The anti-monotone property states that the support for an
itemset never exceeds the support for its subsets. This is the basis for the Apriori
algorithm’s pruning.
6. In K-means clustering, which of the following is a known limitation of the Sum of Squared
Error (SSE) criterion?
A. It favors clusters with non-spherical shapes.
B. It is highly sensitive to outliers.
COMPREHENSIVE QUESTIONS AND
ANSWERS. VERIFIED 2026/2027
1. In the context of data preprocessing, which of the following is an effect of the ‘Curse of
Dimensionality’?
A. Data points become more concentrated in the center of the space.
B. Computational efficiency increases due to the sparsity of the data.
C. The distance between any two points in high-dimensional space becomes nearly
equidistant.
D. Feature selection becomes redundant as all dimensions contribute equally to the
variance.
Answer: C
Conceptual Explanation: As dimensionality increases, the contrast between the distance
to the nearest and farthest neighbor decreases, making distance-based algorithms like KNN
less effective.
2. Which measure is most appropriate for assessing the similarity between two sparse binary
vectors (e.g., transaction data)?
A. Euclidean Distance
,B. Manhattan Distance
C. Pearson Correlation
D. Jaccard Coefficient
Answer: D
Conceptual Explanation: The Jaccard Coefficient is ideal for asymmetric binary attributes
because it ignores ‘0-0’ matches, which are often uninformative in sparse data.
3. What is the primary difference between Principal Component Analysis (PCA) and Linear
Discriminant Analysis (LDA)?
A. PCA is supervised while LDA is unsupervised.
B. LDA reduces dimensionality while PCA only transforms the feature space.
C. PCA seeks to maximize variance, whereas LDA seeks to maximize class separability.
D. PCA handles non-linear relationships whereas LDA is strictly linear.
Answer: C
Conceptual Explanation: PCA is an unsupervised technique that finds directions of
maximum variance. LDA is supervised and finds a projection that maximizes the ratio of
between-class variance to within-class variance.
4. In decision tree induction, what does a Gini Index of 0 indicate for a specific node?
A. The node is a leaf node and is perfectly pure.
B. The node contains an equal distribution of all classes.
, C. The node has reached the maximum allowed depth.
D. The split at this node is considered the worst possible split.
Answer: A
Conceptual Explanation: A Gini Index of 0 means all records in the node belong to a single
class, representing maximum purity.
5. Which of the following describes the ‘Anti-monotone’ property of the Support measure in
association rule mining?
A. Support increases as the number of items in the itemset increases.
B. If an itemset is infrequent, then all of its supersets must be frequent.
C. The confidence of a rule is always less than or equal to its support.
D. If an itemset is frequent, then all of its subsets must also be frequent.
Answer: D
Conceptual Explanation: The anti-monotone property states that the support for an
itemset never exceeds the support for its subsets. This is the basis for the Apriori
algorithm’s pruning.
6. In K-means clustering, which of the following is a known limitation of the Sum of Squared
Error (SSE) criterion?
A. It favors clusters with non-spherical shapes.
B. It is highly sensitive to outliers.
