Society of Actuaries (SOA) PA Practice Exam

What is the main characteristic of K-Means clustering?

1. It forms clusters based on a predefined number of clusters

2. It merges all data points into one cluster

3. It requires labeled data for clustering

4. It randomly assigns clusters without specific criteria

The correct answer is: It forms clusters based on a predefined number of clusters

The main characteristic of K-Means clustering is that it requires the user to specify a predefined number of clusters. This is a fundamental aspect of the algorithm, as it is designed to partition the dataset into a fixed number of groups, or clusters, based on their features. The process begins by randomly initializing the centroids of these clusters and then iteratively assigning data points to the nearest centroid and updating the centroids based on the assigned data points. By defining the number of clusters beforehand, K-Means allows for structured analysis and facilitates the identification of group patterns within the data. Other options present different methods or concepts that do not align with the defining principle of K-Means clustering. For example, merging all data points into one cluster does not facilitate any meaningful grouping and contradicts the concept of clustering, which aims to differentiate between groups. The requirement for labeled data corresponds more to supervised learning techniques, whereas K-Means is an unsupervised method that does not need prior labels. Lastly, the idea that it randomly assigns clusters without specific criteria misrepresents K-Means as it employs distance measures and specific algorithms to assign clusters based on the proximity of data points to the centroids. Thus, the predetermined number of clusters remains the pivotal characteristic of the