Source code for alibi_detect.cd.sklearn.classifier

import logging
import numpy as np
from functools import partial
from typing import Callable, Dict, Optional, Tuple, Union
from sklearn.base import clone, ClassifierMixin
from sklearn.calibration import CalibratedClassifierCV
from sklearn.exceptions import NotFittedError
from sklearn.ensemble import RandomForestClassifier
from alibi_detect.cd.base import BaseClassifierDrift
from alibi_detect.utils.warnings import deprecated_alias
from alibi_detect.utils.frameworks import Framework

logger = logging.getLogger(__name__)


[docs] class ClassifierDriftSklearn(BaseClassifierDrift):
[docs] @deprecated_alias(preprocess_x_ref='preprocess_at_init') def __init__( self, x_ref: np.ndarray, model: ClassifierMixin, p_val: float = .05, x_ref_preprocessed: bool = False, preprocess_at_init: bool = True, update_x_ref: Optional[Dict[str, int]] = None, preprocess_fn: Optional[Callable] = None, preds_type: str = 'probs', binarize_preds: bool = False, train_size: Optional[float] = .75, n_folds: Optional[int] = None, retrain_from_scratch: bool = True, seed: int = 0, use_calibration: bool = False, calibration_kwargs: Optional[dict] = None, use_oob: bool = False, input_shape: Optional[tuple] = None, data_type: Optional[str] = None, ) -> None: """ Classifier-based drift detector. The classifier is trained on a fraction of the combined reference and test data and drift is detected on the remaining data. To use all the data to detect drift, a stratified cross-validation scheme can be chosen. Parameters ---------- x_ref Data used as reference distribution. model Sklearn classification model used for drift detection. p_val p-value used for the significance of the test. x_ref_preprocessed Whether the given reference data `x_ref` has been preprocessed yet. If `x_ref_preprocessed=True`, only the test data `x` will be preprocessed at prediction time. If `x_ref_preprocessed=False`, the reference data will also be preprocessed. preprocess_at_init Whether to preprocess the reference data when the detector is instantiated. Otherwise, the reference data will be preprocessed at prediction time. Only applies if `x_ref_preprocessed=False`. update_x_ref Reference data can optionally be updated to the last n instances seen by the detector or via reservoir sampling with size n. For the former, the parameter equals {'last': n} while for reservoir sampling {'reservoir_sampling': n} is passed. preprocess_fn Function to preprocess the data before computing the data drift metrics. preds_type Whether the model outputs 'probs' or 'scores'. binarize_preds Whether to test for discrepancy on soft (e.g. probs/scores) model predictions directly with a K-S test or binarise to 0-1 prediction errors and apply a binomial test. train_size Optional fraction (float between 0 and 1) of the dataset used to train the classifier. The drift is detected on `1 - train_size`. Cannot be used in combination with `n_folds`. n_folds Optional number of stratified folds used for training. The model preds are then calculated on all the out-of-fold predictions. This allows to leverage all the reference and test data for drift detection at the expense of longer computation. If both `train_size` and `n_folds` are specified, `n_folds` is prioritized. retrain_from_scratch Whether the classifier should be retrained from scratch for each set of test data or whether it should instead continue training from where it left off on the previous set. seed Optional random seed for fold selection. use_calibration Whether to use calibration. Whether to use calibration. Calibration can be used on top of any model. calibration_kwargs Optional additional kwargs for calibration. See https://scikit-learn.org/stable/modules/generated/sklearn.calibration.CalibratedClassifierCV.html for more details. use_oob Whether to use out-of-bag(OOB) predictions. Supported only for `RandomForestClassifier`. input_shape Shape of input data. data_type Optionally specify the data type (tabular, image or time-series). Added to metadata. """ super().__init__( x_ref=x_ref, p_val=p_val, x_ref_preprocessed=x_ref_preprocessed, preprocess_at_init=preprocess_at_init, update_x_ref=update_x_ref, preprocess_fn=preprocess_fn, preds_type=preds_type, binarize_preds=binarize_preds, train_size=train_size, n_folds=n_folds, retrain_from_scratch=retrain_from_scratch, seed=seed, input_shape=input_shape, data_type=data_type ) if preds_type not in ['probs', 'scores']: raise ValueError("'preds_type' should be 'probs' or 'scores'") self.meta.update({'backend': Framework.SKLEARN.value}) self.original_model = model self.use_calibration = use_calibration self.calibration_kwargs = dict() if calibration_kwargs is None else calibration_kwargs self.use_oob = use_oob self.model: ClassifierMixin = self._clone_model()
def _has_predict_proba(self, model) -> bool: try: # taking self.x_ref[0].shape to overcome bot cases when self.x_ref is np.ndarray or list model.predict_proba(np.zeros((1, ) + self.x_ref[0].shape)) has_predict_proba = True except NotFittedError: has_predict_proba = True except AttributeError: has_predict_proba = False return has_predict_proba def _clone_model(self): model = clone(self.original_model) # equivalence between `retrain_from_scratch` and `warm_start` if not self.retrain_from_scratch: if hasattr(model, 'warm_start'): model.warm_start = True logger.warning('`retrain_from_scratch=False` sets automatically the parameter `warm_start=True` ' 'for the given classifier. Please consult the documentation to ensure that the ' '`warm_start=True` is applicable in the current context (i.e., for tree-based ' 'models such as RandomForest, setting `warm_start=True` is not applicable since the ' 'fit function expects the same dataset and an update/increase in the number of ' 'estimators - previous fitted estimators will be kept frozen while the new ones ' 'will be fitted).') else: logger.warning('Current classifier does not support `warm_start`. The model will be retrained ' 'from scratch every iteration.') else: if hasattr(model, 'warm_start'): model.warm_start = False logger.warning('`retrain_from_scratch=True` sets automatically the parameter `warm_start=False`.') # oob checks if self.use_oob: if not isinstance(model, RandomForestClassifier): raise ValueError('OOB supported only for RandomForestClassifier. ' f'Received a model of type {model.__class__.__name__}') if self.use_calibration: self.use_calibration = False logger.warning('Calibration cannot be used when `use_oob=True`. Setting `use_calibration=False`.') model.oob_score = True model.bootstrap = True logger.warning( '`use_oob=True` sets automatically the classifier parameters `boostrap=True` and `oob_score=True`. ' '`train_size` and `n_folds` are ignored when `use_oob=True`.' ) else: if isinstance(model, RandomForestClassifier): model.oob_score = False logger.warning('`use_oob=False` sets automatically the classifier parameters `oob_score=False`.') # preds_type checks if self.preds_type == 'probs': # calibrate the model if user specified. if self.use_calibration: model = CalibratedClassifierCV(estimator=model, **self.calibration_kwargs) logger.warning('Using calibration to obtain the prediction probabilities.') # check if it has predict proba. Cannot be checked via `hasattr` due to the same issue in SVC (see below) has_predict_proba = self._has_predict_proba(model) # if the binarize_preds=True, we don't really need the probabilities as in test_probs will be rounded # to the closest integer (i.e., to 0 or 1) according to the predicted probability. Thus, we can define # a hard label predict_proba based on the predict method if self.binarize_preds and (not has_predict_proba): if not hasattr(model, 'predict'): raise AttributeError('Trying to use a model which does not support `predict`.') def predict_proba(self, X): return np.eye(2)[self.predict(X).astype(np.int32)] # add predict_proba method. Overwriting predict_proba is not possible for SVC due # to @available_if(_check_proba) # Check link: https://github.com/scikit-learn/scikit-learn/blob/7e1e6d09b/sklearn/svm/_base.py#L807 setattr(model, 'aux_predict_proba', partial(predict_proba, model)) elif has_predict_proba: setattr(model, 'aux_predict_proba', model.predict_proba) # at this point the model does not have any predict_proba, thus the test can not be performed. if not hasattr(model, 'aux_predict_proba'): raise AttributeError("Trying to use a model which does not support `predict_proba` with " "`preds_type='probs'`. Set (`use_calibration=True`, `calibration_kwargs`) or " "(`binarize_preds=True`).") else: if self.use_calibration: logger.warning("No calibration is performed when `preds_type='scores'`.") if self.binarize_preds: raise ValueError("`binarize_preds` must be `False` when `preds_type='scores'`.") if not hasattr(model, 'decision_function'): raise AttributeError("Trying to use a model which does not support `decision_function` with " "`preds_type='scores'`.") # need to put the scores in the format expected by test function, which requires to duplicate the # scores along axis=1 def predict_proba(self, X): scores = self.decision_function(X).reshape(-1, 1) return np.tile(scores, reps=2) # add predict_proba method setattr(model, 'aux_predict_proba', partial(predict_proba, model)) return model
[docs] def score(self, x: Union[np.ndarray, list]) \ -> Tuple[float, float, np.ndarray, np.ndarray, Union[np.ndarray, list], Union[np.ndarray, list]]: """ Compute the out-of-fold drift metric such as the accuracy from a classifier trained to distinguish the reference data from the data to be tested. Parameters ---------- x Batch of instances. Returns ------- p-value, a notion of distance between the trained classifier's out-of-fold performance \ and that which we'd expect under the null assumption of no drift, \ and the out-of-fold classifier model prediction probabilities on the reference and test data \ as well as the associated reference and test instances of the out-of-fold predictions. """ if self.use_oob and isinstance(self.model, RandomForestClassifier): return self._score_rf(x) return self._score(x)
def _score(self, x: Union[np.ndarray, list]) \ -> Tuple[float, float, np.ndarray, np.ndarray, Union[np.ndarray, list], Union[np.ndarray, list]]: x_ref, x = self.preprocess(x) x, y, splits = self.get_splits(x_ref, x, return_splits=True) # type: ignore # iterate over folds: train a new model for each fold and make out-of-fold (oof) predictions probs_oof_list, idx_oof_list = [], [] for idx_tr, idx_te in splits: y_tr = y[idx_tr] if isinstance(x, np.ndarray): x_tr, x_te = x[idx_tr], x[idx_te] elif isinstance(x, list): x_tr, x_te = [x[_] for _ in idx_tr], [x[_] for _ in idx_te] else: raise TypeError(f'x needs to be of type np.ndarray or list and not {type(x)}.') self.model.fit(x_tr, y_tr) probs = self.model.aux_predict_proba(x_te) probs_oof_list.append(probs) idx_oof_list.append(idx_te) probs_oof = np.concatenate(probs_oof_list, axis=0) idx_oof = np.concatenate(idx_oof_list, axis=0) y_oof = y[idx_oof] n_cur = y_oof.sum() n_ref = len(y_oof) - n_cur p_val, dist = self.test_probs(y_oof, probs_oof, n_ref, n_cur) idx_sort = np.argsort(idx_oof) probs_sort = probs_oof[idx_sort] if isinstance(x, np.ndarray): x_oof = x[idx_oof] x_sort = x_oof[idx_sort] else: x_oof = [x[_] for _ in idx_oof] x_sort = [x_oof[_] for _ in idx_sort] return p_val, dist, probs_sort[:n_ref, 1], probs_sort[n_ref:, 1], x_sort[:n_ref], x_sort[n_ref:] def _score_rf(self, x: Union[np.ndarray, list]) \ -> Tuple[float, float, np.ndarray, np.ndarray, Union[np.ndarray, list], Union[np.ndarray, list]]: x_ref, x = self.preprocess(x) x, y = self.get_splits(x_ref, x, return_splits=False) # type: ignore self.model.fit(x, y) # it is possible that some inputs do not have OOB scores. This is probably means # that too few trees were used to compute any reliable estimates. idx_oob = np.where(np.all(~np.isnan(self.model.oob_decision_function_), axis=1))[0] probs_oob = self.model.oob_decision_function_[idx_oob] y_oob = y[idx_oob] if isinstance(x, np.ndarray): x_oob: Union[list, np.ndarray] = x[idx_oob] elif isinstance(x, list): x_oob = [x[_] for _ in idx_oob] else: raise TypeError(f'x needs to be of type np.ndarray or list and not {type(x)}.') # comparison due to ordering in get_split (i.e, x = [x_ref, x]) n_ref = np.sum(idx_oob < len(x_ref)).item() n_cur = np.sum(idx_oob >= len(x_ref)).item() p_val, dist = self.test_probs(y_oob, probs_oob, n_ref, n_cur) return p_val, dist, probs_oob[:n_ref, 1], probs_oob[n_ref:, 1], x_oob[:n_ref], x_oob[n_ref:]