K-Means Clustering DATA-SCIENCE-FOUNDATIONS Practice Questions (Page 6)

Questions 26–28

1. 26foundation · easy

   Why can random initialization of centroids lead to different clustering results on the same data?

   • ARandom initialization changes the distance metric.
   • BK-means may converge to different local optima depending on the starting centroids.
   • CRandom initialization changes the number of clusters.
   • DRandom initialization alters the data points themselves.
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A data scientist runs K-means with k=4 on customer purchase data. The algorithm converges to a solution where one cluster contains nearly all customers and the other three clusters each contain only a few outliers. The scientist suspects the random initialization caused this poor result. Which approach is most likely to yield a more balanced and meaningful clustering?

   • AReduce k to 2 to avoid the outlier clusters.
   • BIncrease the maximum number of iterations to allow the algorithm more time to converge.
   • CRun K-means multiple times with different random seeds and select the run with the lowest total within-cluster sum of squares (WCSS).
   • DUse a different distance metric, such as cosine similarity.
   Reveal Solution Discussion Select an answer first
3. 28expert · medium

   A data scientist runs K-means on a large dataset and sets the maximum number of iterations to 50. After 50 iterations, the algorithm stops, but the centroids are still moving noticeably. Which conclusion is most appropriate?

   • AThe algorithm converged because it reached the maximum iterations.
   • BThe algorithm did not converge; the iteration limit was reached before convergence.
   • CThe algorithm converged because the centroids are still moving, which is normal.
   • DThe algorithm will never converge with this dataset.
   Reveal Solution Discussion Select an answer first