K-Means Clustering DATA-SCIENCE-FOUNDATIONS Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   After applying K-means to customer data with k=3, a data scientist examines the cluster centroids. The centroids are: Cluster 1: (age=25, income=$30k), Cluster 2: (age=45, income=$80k), Cluster 3: (age=60, income=$120k). Which interpretation is most valid?

   • ACluster 1 contains only customers aged exactly 25 with income exactly $30k.
   • BCluster 1 represents younger customers with lower income, Cluster 2 represents middle-aged customers with moderate income, and Cluster 3 represents older customers with higher income.
   • CThe clusters are well-separated because the centroids are far apart.
   • DCluster 3 is the most important cluster because it has the highest income.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   What is a consequence of using a distance metric other than Euclidean distance in K-means?

   • AThe clusters become spherical.
   • BThe algorithm always converges faster.
   • CThe centroid may no longer be the mean of the points in the cluster.
   • DThe number of clusters must be increased.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   How does the presence of outliers affect K-means clustering?

   • AOutliers have no effect on the final centroids.
   • BOutliers cause the algorithm to converge faster.
   • COutliers can pull centroids away from the true cluster centers.
   • DOutliers are automatically assigned to their own cluster.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A data scientist is using K-means++ to initialize centroids for a customer segmentation project. Which statement best describes the benefit of K-means++ over random initialization?

   • AIt guarantees the global optimum solution for any dataset.
   • BIt ensures all clusters have the same number of data points.
   • CIt eliminates the need to choose the number of clusters k.
   • DIt spreads out the initial centroids, reducing the chance of poor local optima and often leading to faster convergence.
   Reveal Solution Discussion Select an answer first
6. 15expert · medium

   A data scientist is analyzing a dataset with three natural groups, but one group is much larger and denser than the other two. The scientist runs K-means with k=3 and finds that the large group is split into two, while the two smaller groups are merged into one. Which approach is most likely to produce clusters that match the natural groups?

   • AIncrease k to 4 to allow the large group to be split further.
   • BUse K-means++ initialization to improve the starting centroids.
   • CStandardize the features before running K-means.
   • DUse a clustering algorithm that does not assume equal cluster sizes, such as Gaussian mixture models (GMM).
   Reveal Solution Discussion Select an answer first