Struct Dbscan 

#[non_exhaustive]
pub struct Dbscan { /* private fields */ }

DBSCAN (Density-Based Spatial Clustering of Applications with Noise).

Points are classified as core, border, or noise based on neighborhood density. Supports configurable distance metrics and KD-tree acceleration.

Example

use scry_learn::cluster::Dbscan;
use scry_learn::dataset::Dataset;

let data = Dataset::new(
    vec![vec![0.0, 0.0, 10.0, 10.0], vec![0.0, 0.0, 10.0, 10.0]],
    vec![0.0; 4],
    vec!["x".into(), "y".into()],
    "label",
);

let mut db = Dbscan::new(5.0, 2);
db.fit(&data).unwrap();
assert_eq!(db.n_clusters(), 2);

Implementations

impl Dbscan

pub fn new(eps: f64, min_samples: usize) -> Self

Create a new DBSCAN model.

Arguments

• eps — maximum distance for two points to be considered neighbors.
• min_samples — minimum number of neighbors for a point to be a core point.

pub fn metric(self, m: DistanceMetric) -> Self

Set the distance metric.

Default is DistanceMetric::Euclidean. KD-tree acceleration is only used with Euclidean distance and ≤ 20 features; other metrics always use brute-force.

pub fn fit(&mut self, data: &Dataset) -> Result<()>

Fit the model on a dataset.

Uses KD-tree for Euclidean distance with ≤ 20 features, brute-force otherwise.

pub fn predict(&self, features: &[Vec<f64>]) -> Result<Vec<i32>>

Predict cluster labels for new points.

Each new point is assigned to the cluster of its nearest core point if that core point is within eps. Otherwise the point is labeled noise (-1).

Example

use scry_learn::cluster::Dbscan;
use scry_learn::dataset::Dataset;

let data = Dataset::new(
    vec![vec![0.0, 0.0, 0.0, 10.0, 10.0, 10.0],
         vec![0.0, 0.0, 0.0, 10.0, 10.0, 10.0]],
    vec![0.0; 6],
    vec!["x".into(), "y".into()],
    "label",
);

let mut db = Dbscan::new(5.0, 2);
db.fit(&data).unwrap();

let preds = db.predict(&[vec![0.5, 0.5]]).unwrap();
assert!(preds[0] >= 0, "Should be assigned to a cluster");

pub fn labels(&self) -> &[i32]

Get cluster labels (-1 = noise).

pub fn n_clusters(&self) -> usize

Number of clusters found (excluding noise).

pub fn n_noise(&self) -> usize

Number of noise points.

pub fn n_core_points(&self) -> usize

Number of core points identified during fitting.