Use this clustering method to explore structure in numeric observations after selecting a meaningful feature representation, distance or similarity measure, and algorithm-specific controls.

Points

Required list input.

Clusters

Required integer input.

Random seed

Required integer input.

Choose reproducible initial centroids with k-means++ distance-weighted sampling. The tool evaluates the supplied inputs together and returns the named outputs below; it does not infer omitted operating conditions or change the units shown.^[1]

Initial centroids

The resulting initial centroids returned as a list.

Centroid indices

The resulting centroid indices returned as a list.

• Clusters are model-dependent rather than ground truth. Scaling, outliers, dimensionality, initialization, stopping criteria, density variation, and hyperparameters can materially change the partition.
• Use finite inputs in the displayed units and preserve more precision than the final presentation requires. Independently verify safety-critical, financial, compliance, or production decisions.

Standardize features when appropriate, repeat stochastic runs, inspect cluster stability and diagnostics, and compare another algorithm whose assumptions differ. Validate usefulness against the downstream question rather than cluster count alone.

• Possibilistic C-Means Clustering Tool

  Fit possibilistic typicalities independently for each cluster using a caller-supplied squared-distance scale; unlike fuzzy memberships, rows need not sum to one.

• Fuzzy C-Means Clustering Tool

  Fit fuzzy c-means memberships that sum to one across clusters for each point; fuzzifier must exceed one.

• K-Means Clustering Tool

  Cluster numeric points by iterative nearest-centroid assignment and arithmetic centroid updates. Farthest-point deterministic initialization avoids random results.