PLASTIC

class capymoa.classifier.PLASTIC

Bases: MOAClassifier

PLASTIC classifier.

PLASTIC [1] is an incremental decision tree that restructures the otherwise pruned subtree. PLASTIC improves upon Extremely Fast Decision Trees (EFDT) by not only revisiting previously splits but also trying to maintain as much as possible of the structure once a split is redone. This process is possible because of the decision tree plasticity: one can alter a tree’s structure without affecting its predictions.

>>> from capymoa.classifier import PLASTIC
>>> from capymoa.datasets import ElectricityTiny
>>> from capymoa.evaluation import prequential_evaluation
>>>
>>> stream = ElectricityTiny()
>>> classifier = PLASTIC(stream.get_schema())
>>> results = prequential_evaluation(stream, classifier, max_instances=1000)
>>> print(f"{results['cumulative'].accuracy():.1f}")
84.4

[1]

Heyden, Marco, et al. “Leveraging plasticity in incremental decision trees.” Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer Nature Switzerland, 2024.

__init__(

schema: Schema,

grace_period: int = 200,

reevaluation_period: int = 200,

nominal_estimator: str = 'NominalAttributeClassObserver',

split_criterion: str | SplitCriterion = 'InfoGainSplitCriterion',

split_confidence: float = 1e-07,

tie_threshold: float = 0.05,

tie_threshold_reevaluation: float = 0.05,

rel_min_delta_g: float = 0.5,

binary_splits: bool = False,

leaf_prediction: Literal['MC', 'NB', 'NBA'] = 'NBA',

max_depth: int = 20,

max_branch_length: int = 5,

) → None

Construct PLASTIC classifier.

Parameters:

• grace_period – The number of instances a leaf should observe between split attempts.

• reevaluation_period – The number of instances an internal node should observe between re-evaluation attempts.

• nominal_estimator – Nominal estimator to use.

• split_criterion – Split criterion to use.

• split_confidence – The allowable error in split decision when using fixed confidence. Values closer to 0 will take longer to decide.

• tie_threshold – Threshold below which a split will be forced to break ties.

• tie_threshold_reevaluation – Threshold below which a split will be forced to break ties during reevaluation.

• rel_min_delta_g – Relative minimum information gain to split a tie during reevaluation.

• binary_splits – Only allow binary splits.

• leaf_prediction –

  Leaf prediction to use.

  • MC: Majority class

  • NB: Naive Bayes

  • NBA: Naive Bayes Adaptive

• max_depth – Maximum allowed depth of tree.

• max_branch_length – Maximum allowed length of branches during restructuring.

cli_help()

predict(instance: Instance) → int | None

Predict the label of an instance.

The base implementation calls predict_proba() and returns the label with the highest probability.

Parameters:

instance – The instance to predict the label for.

Returns:

The predicted label or None if the classifier is unable to make a prediction.

predict_proba(

instance,

) → ndarray[tuple[Any, ...], dtype[float64]] | None

Return probability estimates for each label.

Parameters:

instance – The instance to estimate the probabilities for.

Returns:

An array of probabilities for each label or None if the classifier is unable to make a prediction.

train(instance)

Train the classifier with a labeled instance.

Parameters:

instance – The labeled instance to train the classifier with.

random_seed: int

The random seed for reproducibility.

When implementing a classifier ensure random number generators are seeded.

schema: Schema

The schema representing the instances.