Weka Bayesian classifier wrappers.
Probabilistic classifiers from
weka.classifiers.bayes, registered under weka. hub keys.Notes
Requires the optional Weka extra:
pip install 'tuiml[weka]' plus a Java runtime (11+) on PATH.Classes
NaiveBayes — Weka's naive Bayes classifier (hub key weka.NaiveBayes).
Wraps
weka.classifiers.bayes.NaiveBayes. Numeric attributes are modelled with a single Gaussian per class by default; the alternatives below trade that assumption for a kernel or a discretization.
Constructor
__init__( self, use_kernel_estimator: bool = False, use_supervised_discretization: bool = False, nominal_features: Optional[Sequence[int]] = None, options: Optional[Sequence[str]] = None, )
Parameters
use_kernel_estimator
bool
= False
Model numeric attributes with a kernel density estimate rather than a single normal distribution (Weka
-K). Better when an attribute is clearly not Gaussian.
use_supervised_discretization
bool
= False
Discretize numeric attributes with supervised (MDL) binning (Weka
-D). Mutually exclusive with use_kernel_estimator.
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.
Raises
ValueError
If both
use_kernel_estimator and use_supervised_discretization are set: Weka accepts only one numeric-attribute strategy at a time.
python
>>> from tuiml.weka import NaiveBayes
>>> from tuiml.datasets import load_iris
>>> data = load_iris()
>>> clf = NaiveBayes().fit(data.X, data.y)
>>> clf.predict(data.X[:5]).shape
(5,)NaiveBayesMultinomial — multinomial naive Bayes (hub key weka.NaiveBayesMultinomial).
Wraps
weka.classifiers.bayes.NaiveBayesMultinomial. Models each attribute as a count, which is the right assumption for bag-of-words text features.
Constructor
__init__( self, nominal_features: Optional[Sequence[int]] = None, options: Optional[Sequence[str]] = None, )
Parameters
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.
Notes
Expects non-negative counts, not arbitrary real-valued features.
python
>>> from tuiml.weka import NaiveBayesMultinomial
>>> import numpy as np
>>> rng = np.random.default_rng(0)
>>> X = rng.integers(0, 5, size=(60, 4)).astype(float)
>>> y = (X[:, 0] > 2).astype(int)
>>> clf = NaiveBayesMultinomial().fit(X, y)
>>> clf.predict(X[:5]).shape
(5,)BayesNet — Bayesian network classifier (hub key weka.BayesNet).
Wraps
weka.classifiers.bayes.BayesNet. Learns a network structure over the attributes and estimates its conditional probability tables, so unlike naive Bayes it can represent dependencies between attributes.
Constructor
__init__( self, search_algorithm: str = 'weka.classifiers.bayes.net.search.local.K2', estimator: str = 'weka.classifiers.bayes.net.estimate.SimpleEstimator', nominal_features: Optional[Sequence[int]] = None, options: Optional[Sequence[str]] = None, )
Parameters
search_algorithm
str
= "weka.classifiers.bayes.net.search.local.K2"
Fully-qualified Weka class implementing the structure search (Weka
-Q).
estimator
str
= "weka.classifiers.bayes.net.estimate.SimpleEstimator"
Fully-qualified Weka class estimating the conditional probability tables (Weka
-E).
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.
Notes
search_algorithm and estimator take a bare class name: Weka's -Q and -E do not accept options embedded in the value, so per-search or per-estimator settings (such as the estimator's -A smoothing prior) cannot be passed here.
BayesNet requires nominal attributes. Numeric columns must be discretized first — either by passing their indices in nominal_features when they already hold integer codes, or by binning them beforehand with discretization.
python
>>> from tuiml.weka import BayesNet
>>> import numpy as np
>>> rng = np.random.default_rng(0)
>>> X = rng.integers(0, 3, size=(60, 4)).astype(float)
>>> y = (X[:, 0] > 1).astype(int)
>>> clf = BayesNet(nominal_features=range(4)).fit(X, y)
>>> clf.predict(X[:5]).shape
(5,)