DSE Unsupervised Learning: Clustering 4 — Questions and Answers
Question 1: What is the key difference between soft clustering and hard clustering?
- Soft clustering uses distance metrics; hard clustering uses probability
- Soft clustering assigns probabilities of membership to multiple clusters; hard clustering assigns each point to exactly one cluster (Correct answer)
- Soft clustering is faster; hard clustering is more accurate
- Soft clustering works on continuous data; hard clustering works on categorical data
Correct answer: Soft clustering assigns probabilities of membership to multiple clusters; hard clustering assigns each point to exactly one cluster
In soft (fuzzy) clustering like GMM, each point has a fractional membership probability across all clusters, unlike hard clustering where membership is binary.
Question 2: When applying K-Means to text data represented as TF-IDF vectors, which distance metric is typically preferred?
- Euclidean distance
- Manhattan distance
- Cosine similarity (Correct answer)
- Hamming distance
Correct answer: Cosine similarity
Cosine similarity measures the angle between document vectors regardless of magnitude, making it more appropriate for high-dimensional sparse TF-IDF representations.
Question 3: What does the 'elbow' in the K-Means elbow method represent?
- The k value where inertia starts increasing
- The k value where marginal reduction in inertia becomes small (Correct answer)
- The optimal silhouette score
- The point where the algorithm stops converging
Correct answer: The k value where marginal reduction in inertia becomes small
The elbow is where adding more clusters yields diminishing returns in inertia reduction, suggesting that additional clusters explain little additional variance.
Question 4: OPTICS is an extension of DBSCAN that addresses which of its main limitations?
- DBSCAN cannot handle noise points
- DBSCAN requires labeled data for initialization
- DBSCAN struggles with clusters of varying density (Correct answer)
- DBSCAN cannot handle high-dimensional data
Correct answer: DBSCAN struggles with clusters of varying density
OPTICS produces a reachability plot that reveals clusters at multiple density levels, overcoming DBSCAN's single global epsilon that fails when clusters have different densities.
Question 5: A customer segmentation model produces 6 clusters, but cluster 3 has a silhouette score of -0.2 while others average 0.6. What should the analyst do?
- Accept the results since the average is positive
- Try a different k value or merge cluster 3 with its nearest cluster (Correct answer)
- Remove all points in cluster 3 from the dataset
- Switch to a supervised learning approach
Correct answer: Try a different k value or merge cluster 3 with its nearest cluster
A negative silhouette score means points in cluster 3 are on average closer to another cluster than their own, indicating a poor cluster assignment that warrants reconfiguration.
Question 6: In hierarchical clustering, what is the time complexity of the naive algorithm for n data points?
- O(n log n)
- O(n²)
- O(n² log n)
- O(n³) (Correct answer)
Correct answer: O(n³)
Naive agglomerative hierarchical clustering requires O(n²) space for the distance matrix and O(n³) time to repeatedly find and update minimum distances across n-1 merge steps.
Question 7: Which statement best describes the concept of 'cohesion' in cluster evaluation?
- How well-separated different clusters are from each other
- How similar points within the same cluster are to each other (Correct answer)
- The ratio of cluster diameter to inter-cluster distance
- The probability that a random point belongs to the nearest cluster
Correct answer: How similar points within the same cluster are to each other
Cohesion measures intra-cluster compactness — high cohesion means points within a cluster are tightly grouped and similar to one another.
DSE Unsupervised Learning: Clustering 4
What is the key difference between soft clustering and hard clustering?