Struct Dbscan 

pub struct Dbscan<D: DistanceMetric = Euclidean> { /* private fields */ }

DBSCAN clustering algorithm, generic over a distance metric.

The default metric is Euclidean (L2), matching the original behavior where epsilon is compared against Euclidean distance.

use clump::{Dbscan, Clustering, NOISE};

let data = vec![
    vec![0.0f32, 0.0], vec![0.1, 0.0], vec![0.0, 0.1], // cluster
    vec![100.0, 100.0], // outlier
];

let labels = Dbscan::new(0.5, 2).fit_predict(&data).unwrap();
assert_eq!(labels[0], labels[1]); // same cluster
assert_eq!(labels[3], NOISE);     // outlier is noise

Implementations

impl Dbscan<Euclidean>

pub fn new(epsilon: f32, min_pts: usize) -> Self

Create a new DBSCAN clusterer with the default Euclidean distance.

Arguments

• epsilon - Maximum distance between two points to be neighbors.
• min_pts - Minimum number of points to form a dense region.

Typical Values

• epsilon: Often determined by k-distance plot (k = min_pts - 1).
• min_pts: 2 * dimension is a common heuristic. Minimum is 3.

impl<D: DistanceMetric> Dbscan<D>

pub fn with_metric(epsilon: f32, min_pts: usize, metric: D) -> Self

Create a new DBSCAN clusterer with a custom distance metric.

pub fn with_epsilon(self, epsilon: f32) -> Self

Set epsilon (neighborhood radius).

pub fn with_min_pts(self, min_pts: usize) -> Self

Set minimum points for core classification.

pub fn fit_predict_with_noise( &self, data: &[Vec<f32>], ) -> Result<Vec<Option<usize>>>

Fit and predict, returning labels where noise is marked as None.

This is a convenience wrapper so callers don’t need to import DbscanExt.

pub fn is_noise(label: usize) -> bool

Check whether a label is the DBSCAN noise sentinel.

Trait Implementations

impl<D: Clone + DistanceMetric> Clone for Dbscan<D>

fn clone(&self) -> Dbscan<D>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<D: DistanceMetric> Clustering for Dbscan<D>

fn n_clusters(&self) -> usize

DBSCAN discovers clusters dynamically, so this returns 0.

To get the actual number of clusters, examine the labels after fit_predict.

fn fit_predict(&self, data: &[Vec<f32>]) -> Result<Vec<usize>>

Fit the model (if needed) and return one cluster label per input point.

impl<D: DistanceMetric> DbscanExt for Dbscan<D>

fn fit_predict_with_noise( &self, data: &[Vec<f32>], ) -> Result<Vec<Option<usize>>>

Fit and predict, returning labels where noise is marked as None.

fn is_noise(label: usize) -> bool

Check if a label represents noise.

impl<D: Debug + DistanceMetric> Debug for Dbscan<D>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Dbscan<Euclidean>

fn default() -> Self

Returns the “default value” for a type.