In some clustering analyses, the analyst must pre-determine the number of clusters. Which description best matches this requirement?

In some clustering analyses, the analyst must pre-determine the number of clusters. Which description best matches this requirement?

• A. Auto-detects the number of clusters
• B. Requires specific priori
• C. Assumes clusters are perfectly spherical
• D. Is independent of initialization

Answer: (B)

When you must set the number of groups in advance, you’re bringing a priori knowledge about how many clusters you expect to form into the analysis. This happens in methods like k-means, where you specify the number of clusters before running the algorithm, and the results hinge on that chosen number. The description that best matches this requirement is having a priori information about the number of clusters. Auto-detecting the number of clusters would imply the algorithm determines K from the data itself, which isn’t the described scenario. Assuming clusters are perfectly spherical relates to the shape and geometry of clusters, not to whether the number of clusters is pre-specified. Independence from initialization is about whether starting conditions affect the outcome, which again doesn’t address the need to predefine how many clusters there are.

When you must set the number of groups in advance, you’re bringing a priori knowledge about how many clusters you expect to form into the analysis. This happens in methods like k-means, where you specify the number of clusters before running the algorithm, and the results hinge on that chosen number. The description that best matches this requirement is having a priori information about the number of clusters.

Auto-detecting the number of clusters would imply the algorithm determines K from the data itself, which isn’t the described scenario. Assuming clusters are perfectly spherical relates to the shape and geometry of clusters, not to whether the number of clusters is pre-specified. Independence from initialization is about whether starting conditions affect the outcome, which again doesn’t address the need to predefine how many clusters there are.