Clustering GMLE Practice Questions (Page 7)

Questions 31–35

1. 31application · medium

   A data scientist is comparing K-Means and GMM on the same dataset. The dataset has true class labels from a previous study. The scientist wants to know which clustering method produces clusters that better match the known labels. Which metric should they use?

   • AAdjusted Rand Index
   • BSilhouette score
   • CDavies-Bouldin index
   • DInertia
   Reveal Solution Discussion Select an answer first
2. 32expert · hard

   A data scientist is using DBSCAN on a dataset with two features. They set epsilon=0.5 and minPts=5. After running, they find that many points are labeled as noise, but they expected them to form a cluster. What is the most likely reason?

   • AThe distance metric is inappropriate; use a different metric.
   • BThe data needs to be standardized before applying DBSCAN.
   • CThe epsilon is too small, so the density requirement is not met for those points.
   • DThe minPts is too low, causing points to be incorrectly labeled as noise.
   Reveal Solution Discussion Select an answer first
4. 33foundation · easy

   Which statement correctly describes clustering as an unsupervised learning task?

   • AClustering predicts a continuous target variable from input features.
   • BClustering reduces the dimensionality of data while preserving variance.
   • CClustering groups data points based on predefined class labels.
   • DClustering identifies natural groupings in data without using predefined labels.
   Reveal Solution Discussion Select an answer first
5. 34application · medium

   A cybersecurity team is clustering network traffic to detect anomalies. They have a dataset with many normal traffic patterns and a few rare attack patterns. The attack patterns are not dense and are scattered. They want to identify the attack patterns as noise or separate clusters. Which clustering approach is most suitable?

   • AGMM with 2 components
   • BK-Means with k=2
   • CDBSCAN with a small epsilon and minPts=5
   • DHierarchical clustering with complete linkage
   Reveal Solution Discussion Select an answer first
6. 35application · medium

   A data analyst is clustering customer purchase behavior data. They suspect that some customers belong to multiple segments (e.g., both 'frequent buyer' and 'discount seeker'). They want a model that provides a probability of membership for each customer to each segment. Which algorithm should they choose?

   • ADBSCAN, because it can identify noise points that do not belong to any cluster.
   • BK-Means, because it is fast and can be run multiple times with different initializations.
   • CGaussian Mixture Model (GMM), because it provides soft assignments via posterior probabilities.
   • DAgglomerative hierarchical clustering, because it produces a dendrogram that can be cut at different levels.
   Reveal Solution Discussion Select an answer first