Cluster Analysis

Cluster analysis groups items so members of a group are more similar to each other than to members of other groups.

It is unsupervised: it finds structure without pre-labeled examples. Common algorithms include k-means, hierarchical clustering, DBSCAN, and Gaussian mixture models. Each defines closeness differently. Marketing uses clusters to segment customers by purchase behavior for targeted promotions. Genetics groups gene expression profiles to find disease subtypes.

Finance groups assets that move together for diversification. Recommenders cluster user preferences to suggest products. Security tools cluster network logs to spot odd activity. City planners cluster traffic incidents to rank safety work.

Automatic grouping without labels is how organizations make sense of large unlabeled datasets.

You pick a distance and an algorithm. k-means needs k and spherical blobs. Hierarchical clustering builds a tree of merges. DBSCAN finds dense regions and marks outliers. Gaussian mixtures fit overlapping ellipses. None of these need labels. Marketers get segments. Geneticists get subtypes. Traders get correlated assets. Recommenders get user groups.

SOC tools get clusters of logs that are not like the rest. Cities get maps of crash hotspots. The grouping is only as meaningful as the features and the distance you chose. k-means (1967) is still the first clustering method people run: pick k centers, assign points, move centers, repeat.