Source code for evalml.objectives.objective_base

from abc import ABC, abstractmethod

import numpy as np


[docs]class ObjectiveBase(ABC): """Base class for all objectives.""" @property @classmethod @abstractmethod def name(cls): """Returns a name describing the objective.""" @property @classmethod @abstractmethod def greater_is_better(cls): """Returns a boolean determining if a greater score indicates better model performance.""" @property @classmethod @abstractmethod def score_needs_proba(cls): """Returns a boolean determining if the score() method needs probability estimates. This should be true for objectives which work with predicted probabilities, like log loss or AUC, and false for objectives which compare predicted class labels to the actual labels, like F1 or correlation. """
[docs] @classmethod @abstractmethod def objective_function(cls, y_true, y_predicted, X=None): """Computes the relative value of the provided predictions compared to the actual labels, according a specified metric Arguments: y_predicted (pd.Series) : predicted values of length [n_samples] y_true (pd.Series) : actual class labels of length [n_samples] X (pd.DataFrame or np.array) : extra data of shape [n_samples, n_features] necessary to calculate score Returns: numerical value used to calculate score """
[docs] def score(self, y_true, y_predicted, X=None): """Returns a numerical score indicating performance based on the differences between the predicted and actual values. Arguments: y_predicted (pd.Series) : predicted values of length [n_samples] y_true (pd.Series) : actual class labels of length [n_samples] X (pd.DataFrame or np.array) : extra data of shape [n_samples, n_features] necessary to calculate score Returns: score """ self.validate_inputs(y_true, y_predicted) return self.objective_function(y_true, y_predicted, X=X)
[docs] def validate_inputs(self, y_true, y_predicted): """Validates the input based on a few simple checks. Arguments: y_predicted (pd.Series) : predicted values of length [n_samples] y_true (pd.Series) : actual class labels of length [n_samples] Returns: None """ if len(y_predicted) != len(y_true): raise ValueError("Inputs have mismatched dimensions: y_predicted has shape {}, y_true has shape {}".format(len(y_predicted), len(y_true))) if len(y_true) == 0: raise ValueError("Length of inputs is 0") if np.isnan(y_true).any() or np.isinf(y_true).any(): raise ValueError("y_true contains NaN or infinity") if np.isnan(y_predicted).any() or np.isinf(y_predicted).any(): raise ValueError("y_predicted contains NaN or infinity") if self.score_needs_proba and np.any([(y_predicted < 0) | (y_predicted > 1)]): raise ValueError("y_predicted contains probability estimates not within [0, 1]")