Ugz6dZddlmZddlZddlmZddlmZddl m Z m Z ddl m Z dd lmZdd lmZdd lmZdd lmZdd lmZmZmZddlmZmZmZddlmZddl m!Z!m"Z"m#Z#m$Z$m%Z%ddl&m'Z'ddl(m)Z)m*Z*ddl+m,Z,m-Z-dgZ.dZ/dZ0 d-dZ1e%e!ddge!ddge!ddggddgde"e2gddgddgddge$hdge3ge#e"ddd !ge$hd"ge$hd#gd$ d%&dddd'd(d)d'd*d+d,Z4dS).zBPartial dependence plots for regression and classification models.)IterableN)sparse) mquantiles) is_classifier is_regressor)RandomForestRegressor)BaseGradientBoosting)BaseHistGradientBoosting)NotFittedError)DecisionTreeRegressor)Bunch_safe_indexing check_array)_determine_key_type_get_column_indices _safe_assign)check_matplotlib_support) HasMethodsIntegralInterval StrOptionsvalidate_params) cartesian)_check_sample_weightcheck_is_fitted)_check_feature_names_get_feature_indexpartial_dependencecVt|trt|dkrtdt d|Dstd|d|dkrtd|dkrtdg}t |D]\}} t jt||d }n&#t$r}td |d |d }~wwxYw|s|j d|kr|} nvtt||d |d } t j | d| drtdt j | d| d|d} || t||fS)a Generate a grid of points based on the percentiles of X. The grid is a cartesian product between the columns of ``values``. The ith column of ``values`` consists in ``grid_resolution`` equally-spaced points between the percentiles of the jth column of X. If ``grid_resolution`` is bigger than the number of unique values in the j-th column of X or if the feature is a categorical feature (by inspecting `is_categorical`) , then those unique values will be used instead. Parameters ---------- X : array-like of shape (n_samples, n_target_features) The data. percentiles : tuple of float The percentiles which are used to construct the extreme values of the grid. Must be in [0, 1]. is_categorical : list of bool For each feature, tells whether it is categorical or not. If a feature is categorical, then the values used will be the unique ones (i.e. categories) instead of the percentiles. grid_resolution : int The number of equally spaced points to be placed on the grid for each feature. Returns ------- grid : ndarray of shape (n_points, n_target_features) A value for each feature at each point in the grid. ``n_points`` is always ``<= grid_resolution ** X.shape[1]``. values : list of 1d ndarrays The values with which the grid has been created. The size of each array ``values[j]`` is either ``grid_resolution``, or the number of unique values in ``X[:, j]``, whichever is smaller. rz/'percentiles' must be a sequence of 2 elements.c36K|]}d|cxkodkncVdS)rrN).0xs e/var/www/surfInsights/venv3-11/lib/python3.11/site-packages/sklearn/inspection/_partial_dependence.py z_grid_from_X..Ss600qqA{{{{{{{{000000z''percentiles' values must be in [0, 1].rrz9percentiles[0] must be strictly less than percentiles[1].z2'grid_resolution' must be strictly greater than 1.axisz The column #z contains mixed data types. Finding unique categories fail due to sorting. It usually means that the column contains `np.nan` values together with `str` categories. Such use case is not yet supported in scikit-learn.N)probr*ztpercentiles are too close to each other, unable to build the grid. Please choose percentiles that are further apart.T)numendpoint) isinstancerlen ValueErrorall enumeratenpuniquer TypeErrorshaperallcloselinspaceappendr) X percentilesis_categoricalgrid_resolutionvaluesfeatureis_catuniquesexcr*emp_percentiless r& _grid_from_XrD)sP k8 , ,LK0@0@A0E0EJKKK 00K000 0 0DBCCC1~Q''TUUU!MNNN F%^44"" iq' B B BCCGG   =w===      W]1%77DD)q'2221O{?1-q/ABB  . ;""# D  d V  f $$s-$C C5C00C5ct|||}|jdkr|dd}|S)a} Calculate partial dependence via the recursion method. The recursion method is in particular enabled for tree-based estimators. For each `grid` value, a weighted tree traversal is performed: if a split node involves an input feature of interest, the corresponding left or right branch is followed; otherwise both branches are followed, each branch being weighted by the fraction of training samples that entered that branch. Finally, the partial dependence is given by a weighted average of all the visited leaves values. This method is more efficient in terms of speed than the `'brute'` method (:func:`~sklearn.inspection._partial_dependence._partial_dependence_brute`). However, here, the partial dependence computation is done explicitly with the `X` used during training of `est`. Parameters ---------- est : BaseEstimator A fitted estimator object implementing :term:`predict` or :term:`decision_function`. Multioutput-multiclass classifiers are not supported. Note that `'recursion'` is only supported for some tree-based estimators (namely :class:`~sklearn.ensemble.GradientBoostingClassifier`, :class:`~sklearn.ensemble.GradientBoostingRegressor`, :class:`~sklearn.ensemble.HistGradientBoostingClassifier`, :class:`~sklearn.ensemble.HistGradientBoostingRegressor`, :class:`~sklearn.tree.DecisionTreeRegressor`, :class:`~sklearn.ensemble.RandomForestRegressor`, ). grid : array-like of shape (n_points, n_target_features) The grid of feature values for which the partial dependence is calculated. Note that `n_points` is the number of points in the grid and `n_target_features` is the number of features you are doing partial dependence at. features : array-like of {int, str} The feature (e.g. `[0]`) or pair of interacting features (e.g. `[(0, 1)]`) for which the partial dependency should be computed. Returns ------- averaged_predictions : array-like of shape (n_targets, n_points) The averaged predictions for the given `grid` of features values. Note that `n_targets` is the number of targets (e.g. 1 for binary classification, `n_tasks` for multi-output regression, and `n_classes` for multiclass classification) and `n_points` is the number of points in the `grid`. r)%_compute_partial_dependence_recursionndimreshape)estgridfeaturesaveraged_predictionss r&_partial_dependence_recursionrNsGbDDT8TT A%% 4;;ArBB r(cg}g}t|r|j}nrt|dd} t|dd} |dkr| p| }n |dkr| n| }|9|dkrtd|dkrtdtd|} |D]} t |D]\} }t | | | | || }|||tj |d | {#t$r}td |d}~wwxYw|j d }tj |j }t|r"|jd kr||d }n>t!|r/|j d d kr|d}||d }tj |j }t|r"|jdkr|dd }n>t!|r/|j d d kr|d}|dd }||fS)a&Calculate partial dependence via the brute force method. The brute method explicitly averages the predictions of an estimator over a grid of feature values. For each `grid` value, all the samples from `X` have their variables of interest replaced by that specific `grid` value. The predictions are then made and averaged across the samples. This method is slower than the `'recursion'` (:func:`~sklearn.inspection._partial_dependence._partial_dependence_recursion`) version for estimators with this second option. However, with the `'brute'` force method, the average will be done with the given `X` and not the `X` used during training, as it is done in the `'recursion'` version. Therefore the average can always accept `sample_weight` (even when the estimator was fitted without). Parameters ---------- est : BaseEstimator A fitted estimator object implementing :term:`predict`, :term:`predict_proba`, or :term:`decision_function`. Multioutput-multiclass classifiers are not supported. grid : array-like of shape (n_points, n_target_features) The grid of feature values for which the partial dependence is calculated. Note that `n_points` is the number of points in the grid and `n_target_features` is the number of features you are doing partial dependence at. features : array-like of {int, str} The feature (e.g. `[0]`) or pair of interacting features (e.g. `[(0, 1)]`) for which the partial dependency should be computed. X : array-like of shape (n_samples, n_features) `X` is used to generate values for the complement features. That is, for each value in `grid`, the method will average the prediction of each sample from `X` having that grid value for `features`. response_method : {'auto', 'predict_proba', 'decision_function'}, default='auto' Specifies whether to use :term:`predict_proba` or :term:`decision_function` as the target response. For regressors this parameter is ignored and the response is always the output of :term:`predict`. By default, :term:`predict_proba` is tried first and we revert to :term:`decision_function` if it doesn't exist. sample_weight : array-like of shape (n_samples,), default=None Sample weights are used to calculate weighted means when averaging the model output. If `None`, then samples are equally weighted. Note that `sample_weight` does not change the individual predictions. Returns ------- averaged_predictions : array-like of shape (n_targets, n_points) The averaged predictions for the given `grid` of features values. Note that `n_targets` is the number of targets (e.g. 1 for binary classification, `n_tasks` for multi-output regression, and `n_classes` for multiclass classification) and `n_points` is the number of points in the `grid`. predictions : array-like The predictions for the given `grid` of features values over the samples from `X`. For non-multioutput regression and binary classification the shape is `(n_instances, n_points)` and for multi-output regression and multiclass classification the shape is `(n_targets, n_instances, n_points)`, where `n_targets` is the number of targets (`n_tasks` for multi-output regression, and `n_classes` for multiclass classification), `n_instances` is the number of instances in `X`, and `n_points` is the number of points in the `grid`. predict_probaNdecision_functionautozCThe estimator has no predict_proba and no decision_function method.z*The estimator has no predict_proba method.z.The estimator has no decision_function method.)column_indexerr)r*weightsz0'estimator' parameter must be a fitted estimatorrrFr)rpredictgetattrr0copyr2rr9r3averager r6arrayTrHrIr)rJrKrLr:response_method sample_weight predictionsrMprediction_methodrPrQX_eval new_valuesivariableprede n_sampless r&_partial_dependence_bruterfsPKCSK_d;; #C))D99D>fitrUrPrQz array-likez sparse matrix>rRrPrQleft)closed>rRbrute recursion>bothrX individual) estimatorr:rLr\categorical_features feature_namesr[r;r=methodkindT)prefer_skip_nested_validationrR)g?gffffff?drX)r\rorpr[r;r=rqrrct|t|st|stdt|r4t |jdt jrtdt|ds+tj |st|dt}t|r|dkrtd| d kr| d krtd d } | d kr|td| dkrK|d } nFt |tr |jd } n't |tt t"frd } nd } | d krt |ttt t"fs7d} tdd| |dkrd}|dkr"td||t)||}t+|ddkrWt jt j|dr0td|jddz t jt5||t jd} t;||jd} dgt=| z}nt jjj dkr5j!| krtdj!d| dfd| D}nCjj d vrfd!Dfd"| D}ntd#jd$tEtG|| d%|||\}}| d kr`. .. warning:: For :class:`~sklearn.ensemble.GradientBoostingClassifier` and :class:`~sklearn.ensemble.GradientBoostingRegressor`, the `'recursion'` method (used by default) will not account for the `init` predictor of the boosting process. In practice, this will produce the same values as `'brute'` up to a constant offset in the target response, provided that `init` is a constant estimator (which is the default). However, if `init` is not a constant estimator, the partial dependence values are incorrect for `'recursion'` because the offset will be sample-dependent. It is preferable to use the `'brute'` method. Note that this only applies to :class:`~sklearn.ensemble.GradientBoostingClassifier` and :class:`~sklearn.ensemble.GradientBoostingRegressor`, not to :class:`~sklearn.ensemble.HistGradientBoostingClassifier` and :class:`~sklearn.ensemble.HistGradientBoostingRegressor`. Parameters ---------- estimator : BaseEstimator A fitted estimator object implementing :term:`predict`, :term:`predict_proba`, or :term:`decision_function`. Multioutput-multiclass classifiers are not supported. X : {array-like, sparse matrix or dataframe} of shape (n_samples, n_features) ``X`` is used to generate a grid of values for the target ``features`` (where the partial dependence will be evaluated), and also to generate values for the complement features when the `method` is 'brute'. features : array-like of {int, str, bool} or int or str The feature (e.g. `[0]`) or pair of interacting features (e.g. `[(0, 1)]`) for which the partial dependency should be computed. sample_weight : array-like of shape (n_samples,), default=None Sample weights are used to calculate weighted means when averaging the model output. If `None`, then samples are equally weighted. If `sample_weight` is not `None`, then `method` will be set to `'brute'`. Note that `sample_weight` is ignored for `kind='individual'`. .. versionadded:: 1.3 categorical_features : array-like of shape (n_features,) or shape (n_categorical_features,), dtype={bool, int, str}, default=None Indicates the categorical features. - `None`: no feature will be considered categorical; - boolean array-like: boolean mask of shape `(n_features,)` indicating which features are categorical. Thus, this array has the same shape has `X.shape[1]`; - integer or string array-like: integer indices or strings indicating categorical features. .. versionadded:: 1.2 feature_names : array-like of shape (n_features,), dtype=str, default=None Name of each feature; `feature_names[i]` holds the name of the feature with index `i`. By default, the name of the feature corresponds to their numerical index for NumPy array and their column name for pandas dataframe. .. versionadded:: 1.2 response_method : {'auto', 'predict_proba', 'decision_function'}, default='auto' Specifies whether to use :term:`predict_proba` or :term:`decision_function` as the target response. For regressors this parameter is ignored and the response is always the output of :term:`predict`. By default, :term:`predict_proba` is tried first and we revert to :term:`decision_function` if it doesn't exist. If ``method`` is 'recursion', the response is always the output of :term:`decision_function`. percentiles : tuple of float, default=(0.05, 0.95) The lower and upper percentile used to create the extreme values for the grid. Must be in [0, 1]. grid_resolution : int, default=100 The number of equally spaced points on the grid, for each target feature. method : {'auto', 'recursion', 'brute'}, default='auto' The method used to calculate the averaged predictions: - `'recursion'` is only supported for some tree-based estimators (namely :class:`~sklearn.ensemble.GradientBoostingClassifier`, :class:`~sklearn.ensemble.GradientBoostingRegressor`, :class:`~sklearn.ensemble.HistGradientBoostingClassifier`, :class:`~sklearn.ensemble.HistGradientBoostingRegressor`, :class:`~sklearn.tree.DecisionTreeRegressor`, :class:`~sklearn.ensemble.RandomForestRegressor`, ) when `kind='average'`. This is more efficient in terms of speed. With this method, the target response of a classifier is always the decision function, not the predicted probabilities. Since the `'recursion'` method implicitly computes the average of the Individual Conditional Expectation (ICE) by design, it is not compatible with ICE and thus `kind` must be `'average'`. - `'brute'` is supported for any estimator, but is more computationally intensive. - `'auto'`: the `'recursion'` is used for estimators that support it, and `'brute'` is used otherwise. If `sample_weight` is not `None`, then `'brute'` is used regardless of the estimator. Please see :ref:`this note ` for differences between the `'brute'` and `'recursion'` method. kind : {'average', 'individual', 'both'}, default='average' Whether to return the partial dependence averaged across all the samples in the dataset or one value per sample or both. See Returns below. Note that the fast `method='recursion'` option is only available for `kind='average'` and `sample_weights=None`. Computing individual dependencies and doing weighted averages requires using the slower `method='brute'`. .. versionadded:: 0.24 Returns ------- predictions : :class:`~sklearn.utils.Bunch` Dictionary-like object, with the following attributes. individual : ndarray of shape (n_outputs, n_instances, len(values[0]), len(values[1]), ...) The predictions for all the points in the grid for all samples in X. This is also known as Individual Conditional Expectation (ICE). Only available when `kind='individual'` or `kind='both'`. average : ndarray of shape (n_outputs, len(values[0]), len(values[1]), ...) The predictions for all the points in the grid, averaged over all samples in X (or over the training data if `method` is 'recursion'). Only available when `kind='average'` or `kind='both'`. grid_values : seq of 1d ndarrays The values with which the grid has been created. The generated grid is a cartesian product of the arrays in `grid_values` where `len(grid_values) == len(features)`. The size of each array `grid_values[j]` is either `grid_resolution`, or the number of unique values in `X[:, j]`, whichever is smaller. .. versionadded:: 1.3 `n_outputs` corresponds to the number of classes in a multi-class setting, or to the number of tasks for multi-output regression. For classical regression and binary classification `n_outputs==1`. `n_values_feature_j` corresponds to the size `grid_values[j]`. See Also -------- PartialDependenceDisplay.from_estimator : Plot Partial Dependence. PartialDependenceDisplay : Partial Dependence visualization. Examples -------- >>> X = [[0, 0, 2], [1, 0, 0]] >>> y = [0, 1] >>> from sklearn.ensemble import GradientBoostingClassifier >>> gb = GradientBoostingClassifier(random_state=0).fit(X, y) >>> partial_dependence(gb, features=[0], X=X, percentiles=(0, 1), ... grid_resolution=2) # doctest: +SKIP (array([[-4.52..., 4.52...]]), [array([ 0., 1.])]) z5'estimator' must be a fitted regressor or classifier.rz3Multiclass-multioutput estimators are not supported __array__z allow-nan)force_all_finitedtyperRzKThe response_method parameter is ignored for regressors and must be 'auto'.rXrkzCThe 'recursion' method only applies when 'kind' is set to 'average'rjNzFThe 'recursion' method can only be applied when sample_weight is None.)GradientBoostingClassifierGradientBoostingRegressorHistGradientBoostingClassifierHistGradientBoostingRegressorr|r r z[Only the following estimators support the 'recursion' method: {}. Try using method='brute'.z, rQzUWith the 'recursion' method, the response_method must be 'decision_function'. Got {}.F) accept_sliceintzall features must be in [0, {}]rC)rxorderbzeWhen `categorical_features` is a boolean array-like, the array should be of shape (n_features,). Got z elements while `X` contains z features.c g|] }| Sr#r#)r$idxros r& z&partial_dependence..sTTTC237TTTr()raOUc2g|]}t|S))rp)r)r$catrps r&rz&partial_dependence..s6(((#3mDDD(((r(cg|]}|vSr#r#)r$rcategorical_features_idxs r&rz&partial_dependence..s,47//r(zXExpected `categorical_features` to be an array-like of boolean, integer, or string. Got z instead.r)rFc(g|]}|jdSrr6r$vals r&rz&partial_dependence..s===scil===r(c(g|]}|jdSrrrs r&rz&partial_dependence..s - - -scil - - -r() grid_valuesrm)(rrrr0r.classes_r3ndarrayhasattrrissparserobjectr initr r r formatjoinrranylessr6asarrayrintpravelrr/rxrrsizerDrrfrIrNr)rnr:rLr\rorpr[r;r=rqrrsupported_classes_recursionfeatures_indices n_featuresr<rKr>rMr] pdp_resultsrs `` @r&r r [sjI ) $ $R Y(?(?RPQQQYPJy/A!/Dbj$Q$QPNOOO A{ # #Gvq'9'9G Kv F F FI ?f#<#<     y [ U  !: T     $FF  #7 8 8 Y^=S FF   %'<>S T   !FFF  $(%%     + '88>II9::99  f $ $1O 1 1 1//5vo/F/F   ,]A>> 8%888EAA 6"'(A&& ' ' W>EEagajSTnUUVV VzAx((s egg)M::MJ#3'7#8#88!z*>??  % *c 1 1#(J66 .+0.."... UTTTCSTTTNN ! ' , ? ?((((/((( $;KNNR,@,FRRR   q*333 LD&,E t-q/=- - )k *k)   ==f===    = t-   8/7  - -f - - -F+++K y!5 I   $/ L!!!5 I$/ L! r()N)5__doc__collections.abcrnumpyr3scipyrscipy.stats.mstatsrbaserrensembler ensemble._gbr 2ensemble._hist_gradient_boosting.gradient_boostingr exceptionsr treer utilsrrrutils._indexingrrrutils._optional_dependenciesrutils._param_validationrrrrr utils.extmathrutils.validationrr _pd_utilsrr__all__rDrNrfstrtupler r#r(r&rs5HH%$$$$$))))))........,,,,,,//////('''''((((((6666666666TTTTTTTTTTCCCCCC&%%%%%DDDDDDDD????????  Y%Y%Y%x7 7 7 v<@Y-Y-Y-Y-x Jy) * * J/ 0 0 J23 4 4 O ,!8S1&-!-t 4&-&J'U'U'UVVWw$HXq$vFFFG:<<<==>===>>?"#'%2  bbbb)(bbbr(