OPTICS
Density-based clustering robust to varying cluster densities
OPTICS (Ordering Points To Identify Clustering Structure) is an extension of DBSCAN that handles clusters of varying densities. It produces a reachability plot that can be cut at different thresholds to produce different cluster structures.
When to use:
- Datasets with clusters of significantly different densities
- When DBSCAN produces poor results due to a single global eps parameter
- Exploratory clustering where multiple granularities are of interest
Input: Tabular data with the feature columns defined during training Output: Cluster label per row (-1 for noise)
Model Settings (set during training, used at inference)
Min Samples (default: 5) Minimum neighbors for a core point. Controls minimum cluster size.
Max Eps (default: infinity) Maximum neighborhood radius. Setting this can speed up computation.
Metric (default: minkowski) Distance metric for neighborhood search.
Cluster Method (default: xi)
Method for extracting flat clusters. xi uses the reachability plot gradient; dbscan cuts at a fixed eps value.
Inference Settings
No dedicated inference-time settings. New points are assigned to the nearest core point's cluster.