Clustering Algorithms

The most widely-used clustering algorithm. It partitions data into k clusters by iteratively assigning points to the nearest centroid and updating centroids until convergence.

Similar to K-Means but uses actual data points (medoids) as cluster centers instead of means. More robust to outliers and noise in the data.

A non-parametric algorithm that doesn't require specifying the number of clusters. It finds clusters by shifting points towards the mode (highest density) of the data.

Unlike hard clustering, FCM allows data points to belong to multiple clusters with varying degrees of membership. Perfect for overlapping or ambiguous cluster boundaries.

Discovers clusters of arbitrary shape based on density. Can identify noise points and doesn't require specifying the number of clusters beforehand.

Hierarchical extension of DBSCAN that handles varying density clusters. Currently the gold standard for density-based clustering.

Creates an ordering of the database that captures density-based clustering structure. Produces reachability plots for visual cluster analysis.

Builds a hierarchy of clusters by progressively merging the closest pairs. Produces a dendrogram that can be cut at any level.

Balanced Iterative Reducing and Clustering using Hierarchies. Designed for very large datasets with limited memory.

A divisive (top-down) hierarchical clustering algorithm that repeatedly bisects clusters using K-Means until the desired number of clusters is reached.

Models data as a mixture of Gaussian distributions. Provides soft cluster assignments and can model elliptical clusters of different sizes.

Uses eigenvalues of a similarity matrix to reduce dimensionality before clustering. Excellent for non-convex cluster shapes.

Identifies exemplars among data points by passing messages between pairs. Automatically determines the number of clusters.

A neural network approach that grows nodes to represent data topology. Adapts structure during training to match data distribution.