BaseStructureScore

class pgmpy.structure_score.BaseStructureScore(data, state_names=None)

Bases: BaseObject

Abstract base class for structure scoring in pgmpy.

Structure scores evaluate how well a candidate Bayesian network structure fits observed data. This class implements the shared scoring workflow, caching for local scores, and utilities for computing conditional state counts. Use one of the concrete score classes such as K2, BDeu, BIC, or AIC instead of instantiating this class directly.

Parameters:

datapandas.DataFrame

DataFrame in which each column represents a variable. Missing values should be marked as numpy.nan.

state_namesdict, optional

Dictionary mapping each variable name to its allowed states. If not specified, the observed values in the data are used.

local_score(variable: str, parents: tuple[str, ...]) → float

Compute the cached local score for variable given parents.

score(model) → float

Compute a structure score for a model.

state_counts(variable: str, parents: tuple[str, ...] = (), weighted: bool = False, reindex: bool = True) → DataFrame

Return state counts for variable, optionally conditioned on parents.

structure_prior(model) → float

Return the log prior over structures.

structure_prior_ratio(operation) → float

Return the log prior ratio for a structure operation.