Clustering algorithms group objects such that similar feature points are put into the same groups (clusters).
Proposed by @Al_Grigor
Card 1 of 259. Answer: Clustering algorithms group objects such that similar feature points are put into the same groups (clusters).
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
Clustering is the unsupervised task of grouping data points so that points in the same cluster are more similar to each other than to points in other clusters. There are no labels: the algorithm finds the groups from the features alone, usually with a distance such as Euclidean distance defining similarity.
Example: six customers spend 1, 2, 3, 10, 11 and 12 thousand dollars a year, and k-means with k = 2 finds {1, 2, 3} and {10, 11, 12}. Add a customer who spends 40: k-means, forced to make two clusters, does best by isolating 40 and merging the real groups. DBSCAN, a density-based method, with a neighborhood radius of 1.5 and a minimum of 2 points keeps both groups and labels 40 as noise, without being told how many clusters exist.
Other families include hierarchical clustering and Gaussian mixtures, which give soft memberships. Interviewers often ask how to judge clusters without labels: the silhouette score, plus whether the groups mean something for the business. Scale features first, or the widest-ranging one dominates the distances.
