API Reference / weka /

clusterers.py

Weka clusterer wrappers.

Unsupervised learners from weka.clusterers, registered under weka. hub keys.

Notes

Requires the optional Weka extra: pip install 'tuiml[weka]' plus a Java runtime (11+) on PATH.

Classes

SimpleKMeans

class weka.clusterers.SimpleKMeans(_WekaClustererMixin, Clusterer)

SimpleKMeans — k-means clustering (hub key weka.SimpleKMeans).

Wraps weka.clusterers.SimpleKMeans.
Constructor 
__init__(
    self,
    n_clusters: int = 2,
    max_iterations: int = 500,
    seed: int = 10,
    nominal_features: Optional[Sequence[int]] = None,
    options: Optional[Sequence[str]] = None,
)

Parameters

n_clusters
int = 2
Number of clusters (Weka -N).
max_iterations
int = 500
Maximum iterations (Weka -I).
seed
int = 10
Random seed (Weka -S).
nominal_features
sequence of int or None = None
0-based indices of columns to treat as categorical.
options
sequence of str or None = None
Extra raw Weka options.

Attributes

labels_
np.ndarray of shape (n_samples,)
Cluster assignment for each training row.
n_clusters_
int
Number of clusters produced.

See Also

KMeansClusterer

TuiML's native k-means.
python 
>>> from tuiml.weka import SimpleKMeans
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> km = SimpleKMeans(n_clusters=3, seed=42).fit(data.X)
>>> km.labels_.shape
(150,)

Methods

get_parameter_schema (cls) -> Dict[str, Any]

Return JSON Schema for constructor parameters.

get_capabilities (cls) -> List[str]

Return the capability names this wrapper supports.

EM

class weka.clusterers.EM(_WekaClustererMixin, Clusterer)

EM — expectation-maximization clustering (hub key weka.EM).

Wraps weka.clusterers.EM. Fits a mixture of Gaussians and can choose the number of clusters itself by cross-validation.
Constructor 
__init__(
    self,
    n_clusters: int = ...,
    max_iterations: int = 100,
    seed: int = 100,
    nominal_features: Optional[Sequence[int]] = None,
    options: Optional[Sequence[str]] = None,
)

Parameters

n_clusters
int = -1
Number of clusters (Weka -N); -1 selects it by cross-validation.
max_iterations
int = 100
Maximum EM iterations (Weka -I).
seed
int = 100
Random seed (Weka -S).
nominal_features
sequence of int or None = None
0-based indices of columns to treat as categorical.
options
sequence of str or None = None
Extra raw Weka options.

See Also

GaussianMixtureClusterer

TuiML's native version.
python 
>>> from tuiml.weka import EM
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> em = EM(n_clusters=3, seed=42).fit(data.X)
>>> em.n_clusters_
3

Methods

get_parameter_schema (cls) -> Dict[str, Any]

Return JSON Schema for constructor parameters.

get_capabilities (cls) -> List[str]

Return the capability names this wrapper supports.

Canopy

class weka.clusterers.Canopy(_WekaClustererMixin, Clusterer)

Canopy — cheap approximate pre-clustering (hub key weka.Canopy).

Wraps weka.clusterers.Canopy. Uses two distance thresholds to form overlapping "canopies" in a single cheap pass, normally as a seeding step for a more expensive algorithm.
Constructor 
__init__(
    self,
    n_clusters: int = ...,
    t2: float = ...,
    t1: float = ...,
    seed: int = 1,
    nominal_features: Optional[Sequence[int]] = None,
    options: Optional[Sequence[str]] = None,
)

Parameters

n_clusters
int = -1
Requested number of clusters (Weka -N); -1 lets the thresholds decide.
t2
float = -1.0
Loose distance threshold (Weka -t2); a negative value asks Weka to derive it from the data.
t1
float = -1.25
Tight distance threshold (Weka -t1); a negative value is interpreted as a multiple of t2.
seed
int = 1
Random seed (Weka -S).
nominal_features
sequence of int or None = None
0-based indices of columns to treat as categorical.
options
sequence of str or None = None
Extra raw Weka options.

References

McCallum2000
McCallum, A., Nigam, K. and Ungar, L.H. (2000). Efficient Clustering of High-Dimensional Data Sets with Application to Reference Matching. KDD '00, 169-178.
python 
>>> from tuiml.weka import Canopy
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> cp = Canopy(n_clusters=3, seed=42).fit(data.X)
>>> cp.labels_.shape
(150,)

Methods

get_parameter_schema (cls) -> Dict[str, Any]

Return JSON Schema for constructor parameters.

get_capabilities (cls) -> List[str]

Return the capability names this wrapper supports.

FarthestFirst

class weka.clusterers.FarthestFirst(_WekaClustererMixin, Clusterer)

FarthestFirst — farthest-first traversal clustering (hub key weka.FarthestFirst).

Wraps weka.clusterers.FarthestFirst. Picks each new centre as the point farthest from the existing ones, which is fast and gives well-separated seeds for k-means.
Constructor 
__init__(
    self,
    n_clusters: int = 2,
    seed: int = 1,
    nominal_features: Optional[Sequence[int]] = None,
    options: Optional[Sequence[str]] = None,
)

Parameters

n_clusters
int = 2
Number of clusters (Weka -N).
seed
int = 1
Random seed (Weka -S).
nominal_features
sequence of int or None = None
0-based indices of columns to treat as categorical.
options
sequence of str or None = None
Extra raw Weka options.

See Also

KMeansClusterer

TuiML's native centroid-based clusterer.
python 
>>> from tuiml.weka import FarthestFirst
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> ff = FarthestFirst(n_clusters=3, seed=42).fit(data.X)
>>> ff.labels_.shape
(150,)

Methods

get_parameter_schema (cls) -> Dict[str, Any]

Return JSON Schema for constructor parameters.

get_capabilities (cls) -> List[str]

Return the capability names this wrapper supports.

Cobweb

class weka.clusterers.Cobweb(_WekaClustererMixin, Clusterer)

Cobweb — incremental conceptual clustering (hub key weka.Cobweb).

Wraps weka.clusterers.Cobweb. Builds a hierarchy one instance at a time, guided by category utility. The number of clusters emerges from the data rather than being requested.
Constructor 
__init__(
    self,
    acuity: float = 1.0,
    cutoff: float = 0.002,
    seed: int = 42,
    nominal_features: Optional[Sequence[int]] = None,
    options: Optional[Sequence[str]] = None,
)

Parameters

acuity
float = 1.0
Minimum standard deviation allowed for a numeric attribute (Weka -A).
cutoff
float = 0.002
Category utility threshold below which a node is pruned (Weka -C).
seed
int = 42
Random seed (Weka -S).
nominal_features
sequence of int or None = None
0-based indices of columns to treat as categorical.
options
sequence of str or None = None
Extra raw Weka options.

See Also

AgglomerativeClusterer

TuiML's native hierarchical clusterer.
python 
>>> from tuiml.weka import Cobweb
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> cw = Cobweb().fit(data.X)
>>> cw.labels_.shape
(150,)

Methods

get_parameter_schema (cls) -> Dict[str, Any]

Return JSON Schema for constructor parameters.

get_capabilities (cls) -> List[str]

Return the capability names this wrapper supports.

HierarchicalClusterer

class weka.clusterers.HierarchicalClusterer(_WekaClustererMixin, Clusterer)

HierarchicalClusterer — agglomerative clustering (hub key weka.HierarchicalClusterer).

Wraps weka.clusterers.HierarchicalClusterer. Merges the closest pair of clusters repeatedly under the chosen linkage until the requested number of clusters remains.
Constructor 
__init__(
    self,
    n_clusters: int = 2,
    link_type: str = 'SINGLE',
    nominal_features: Optional[Sequence[int]] = None,
    options: Optional[Sequence[str]] = None,
)

Parameters

n_clusters
int = 2
Number of clusters (Weka -N).
link_type
str = "SINGLE"
Linkage criterion (Weka -L): one of SINGLE, COMPLETE, AVERAGE, MEAN, CENTROID, WARD, ADJCOMPLETE, NEIGHBOR_JOINING.
nominal_features
sequence of int or None = None
0-based indices of columns to treat as categorical.
options
sequence of str or None = None
Extra raw Weka options.

See Also

AgglomerativeClusterer

TuiML's native version.
python 
>>> from tuiml.weka import HierarchicalClusterer
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> hc = HierarchicalClusterer(n_clusters=3, link_type="COMPLETE").fit(data.X)
>>> hc.labels_.shape
(150,)

Methods

get_parameter_schema (cls) -> Dict[str, Any]

Return JSON Schema for constructor parameters.

get_capabilities (cls) -> List[str]

Return the capability names this wrapper supports.