\c@ s=dZddlmZddlmZddlmZddlmZddlmZdd lm Z dd lm Z dd l m Z dd l mZdd l mZddl mZddlZddlZddlmZddlmZddlmZddlmZddlmZddlmZddlmZddlm Z ddlm!Z!ddl"m#Z#ddl"m$Z$ddl"m%Z%ddl"m&Z&ddl"m'Z'ddl"m(Z(ddl)m*Z*dd l+m,Z,dd!l-m.Z.dd"l/m0Z0dd#l1m2Z2d$e3fd%YZ4d&e3fd'YZ5d(e3fd)YZ6d*e6fd+YZ7d,e3fd-YZ8d.e3fd/YZ9d0e j:ee3fd1YZ;d2e j:ee;fd3YZ<d4e<fd5YZ=d6e<fd7YZ>d8e<fd9YZ?d:e<fd;YZ@d<e j:ee;fd=YZAd>eAfd?YZBd@eAfdAYZCdBeAfdCYZDie=dD6e>dE6e?dF6e@dG6eEdH6eDdI6ZFie9dJ6ZGdKe3fdLYZHdMe j:eefdNYZIdOeIe fdPYZJdQeIe fdRYZKdS(Ss>Gradient Boosted Regression Trees This module contains methods for fitting gradient boosted regression trees for both classification and regression. The module structure is the following: - The ``BaseGradientBoosting`` base class implements a common ``fit`` method for all the estimators in the module. Regression and classification only differ in the concrete ``LossFunction`` used. - ``GradientBoostingClassifier`` implements gradient boosting for classification problems. - ``GradientBoostingRegressor`` implements gradient boosting for regression problems. i(tprint_function(tdivision(tABCMeta(tabstractmethodi(t BaseEnsemblei(tClassifierMixin(tRegressorMixin(tsix(tpredict_stages(t predict_stage(t_random_sample_maskN(t csc_matrix(t csr_matrix(tissparse(texpit(ttime(ttrain_test_split(tDecisionTreeRegressor(tDTYPE(t TREE_LEAF(tcheck_random_state(t check_array(t check_X_y(t column_or_1d(tcheck_consistent_length(t deprecated(t logsumexp(t_weighted_percentile(tcheck_is_fitted(tcheck_classification_targets(tNotFittedErrortQuantileEstimatorcB s/eZdZddZddZdZRS(sAn estimator predicting the alpha-quantile of the training targets. Parameters ---------- alpha : float The quantile g?cC s<d|kodkns/td|n||_dS(Nig?s&`alpha` must be in (0, 1.0) but was %r(t ValueErrortalpha(tselfR!((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt__init__IscC sK|dkr+tj||jd|_nt|||jd|_dS(sFit the estimator. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Training data y : array, shape (n_samples, n_targets) Target values. Will be cast to X's dtype if necessary sample_weight : numpy array of shape (n_samples,) Individual weights for each sample gY@N(tNonetnpt percentileR!tquantileR(R"tXtyt sample_weight((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytfitNs  cC sFt|dtj|jddfdtj}|j|j|S(sPredict labels Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Samples. Returns ------- y : array, shape (n_samples,) Returns predicted values. R'iitdtype(RR%temptytshapetfloat64tfillR'(R"R(R)((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytpredictbs %N(t__name__t __module__t__doc__R#R$R+R1(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRAs  t MeanEstimatorcB s#eZdZddZdZRS(s9An estimator predicting the mean of the training targets.cC s=|dkr!tj||_ntj|d||_dS(sFit the estimator. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Training data y : array, shape (n_samples, n_targets) Target values. Will be cast to X's dtype if necessary sample_weight : numpy array of shape (n_samples,) Individual weights for each sample tweightsN(R$R%tmeantaverage(R"R(R)R*((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR+xs cC sFt|dtj|jddfdtj}|j|j|S(sPredict labels Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Samples. Returns ------- y : array, shape (n_samples,) Returns predicted values. R7iiR,(RR%R-R.R/R0R7(R"R(R)((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR1s %N(R2R3R4R$R+R1(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR5vs tLogOddsEstimatorcB s)eZdZdZddZdZRS(s+An estimator predicting the log odds ratio.g?cC s|dkr/tj|}|jd|}n*tj||}tj|d|}|dksq|dkrtdn|jtj|||_dS(sFit the estimator. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Training data y : array, shape (n_samples, n_targets) Target values. Will be cast to X's dtype if necessary sample_weight : numpy array of shape (n_samples,) Individual weights for each sample iisy contains non binary labels.N(R$R%tsumR.R tscaletlogtprior(R"R(R)R*tpostneg((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR+s cC sFt|dtj|jddfdtj}|j|j|S(sPredict labels Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Samples. Returns ------- y : array, shape (n_samples,) Returns predicted values. R=iiR,(RR%R-R.R/R0R=(R"R(R)((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR1s %N(R2R3R4R;R$R+R1(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR9s tScaledLogOddsEstimatorcB seZdZdZRS(s6Log odds ratio scaled by 0.5 -- for exponential loss. g?(R2R3R4R;(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR@stPriorProbabilityEstimatorcB s#eZdZddZdZRS(sTAn estimator predicting the probability of each class in the training data. cC sS|dkr'tj|dtj}ntj|d|}||j|_dS(sxFit the estimator. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Training data y : array, shape (n_samples, n_targets) Target values. Will be cast to X's dtype if necessary sample_weight : array, shape (n_samples,) Individual weights for each sample R,R6N(R$R%t ones_likeR/tbincountR:tpriors(R"R(R)R*t class_counts((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR+s cC sJt|dtj|jd|jjdfdtj}|j|(|S(sPredict labels Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Samples. Returns ------- y : array, shape (n_samples,) Returns predicted values. RDiR,(RR%R-R.RDR/(R"R(R)((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR1s / N(R2R3R4R$R+R1(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRAs t ZeroEstimatorcB s#eZdZddZdZRS(s(An estimator that simply predicts zero. cC s\tj|jtjrOtj|jd|_|jdkrXd|_qXn d|_dS(sxFit the estimator. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Training data y : numpy, shape (n_samples, n_targets) Target values. Will be cast to X's dtype if necessary sample_weight : array, shape (n_samples,) Individual weights for each sample iiiN(R%t issubdtypeR,t signedintegertuniqueR.t n_classes(R"R(R)R*((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR+s cC sFt|dtj|jd|jfdtj}|jd|S(sPredict labels Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) Samples. Returns ------- y : array, shape (n_samples,) Returns predicted values. RJiR,g(RR%R-R.RJR/R0(R"R(R)((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR1s ( N(R2R3R4R$R+R1(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRFs t LossFunctioncB seeZdZeZdZdZed dZ edZ dddZ edZ RS( sLAbstract base class for various loss functions. Parameters ---------- n_classes : int Number of classes Attributes ---------- K : int The number of regression trees to be induced; 1 for regression and binary classification; ``n_classes`` for multi-class classification. cC s ||_dS(N(tK(R"RJ((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR#AscC s tdS(s.Default ``init`` estimator for loss function. N(tNotImplementedError(R"((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytinit_estimatorDscC sdS(s$Compute the loss. Parameters ---------- y : array, shape (n_samples,) True labels pred : array, shape (n_samples,) Predicted labels sample_weight : array-like, shape (n_samples,), optional Sample weights. N((R"R)tpredR*((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt__call__HscK sdS(sCompute the negative gradient. Parameters ---------- y : array, shape (n_samples,) The target labels. y_pred : array, shape (n_samples,) The predictions. N((R"R)ty_predtkargs((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytnegative_gradientXsg?ic C s|j|} | j} d| | 0.0 else -1.0 Parameters ---------- y : array, shape (n_samples,) The target labels. pred : array, shape (n_samples,) The predictions. g@gg?(Rj(R"R)RORR((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRSs c C stj||kd} |j| dd}|j| dd|j| dd} t| |dd|j|ddf10d}s{train_score:>16.4f}is OOB Improves{oob_impr:>16.4f}sRemaining Times{remaining_time:>16s}s%10s s%16s t N( t subsampletappendtprinttlenttupletjoint verbose_fmtt verbose_modRt start_timetbegin_at_stage(R"testRt header_fieldsR((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytinit%s        c C s#|jdk}||j}|d|jdkr|rF|j|nd}|j|dt|jt|d}|dkrdj|d}ndj|}t |j jd|dd|j |d |d ||j dkr|d|jd dkr|jd 9_qnd S( sUpdate reporter with new iteration. Parameters ---------- j : int The new iteration est : Estimator The estimator iii<s{0:.2f}mgN@s{0:.2f}stitert train_scoretoob_imprtremaining_timei N( RRRtoob_improvement_t n_estimatorsRRtfloatRRRt train_score_R(R"tjRtdo_oobtiRR((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytupdateDs  -   *(R2R3R4R#RR(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRs   tBaseGradientBoostingc B seZdZedddedddddZdddZdZd Z d Z d Z d Z d Z eeddZdddZddddZedZdZdZdZedZdZdZRS(s+Abstract base class for Gradient Boosting. g?itautog?g-C6?cC s||_||_||_||_||_||_||_| |_| |_||_ | |_ | |_ | |_ ||_ ||_||_||_||_||_||_||_||_dS(N(RR_Rtt criteriontmin_samples_splittmin_samples_leaftmin_weight_fraction_leafRt max_featurest max_depthtmin_impurity_decreasetmin_impurity_splitRt random_stateR!Rtmax_leaf_nodest warm_starttpresorttvalidation_fractiontn_iter_no_changettol(R"RtR_RRRRRRRRRRRRR!RRRRRRR((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR#es,                     c C s|jtjkst|j} |} xqt| jD]`} | jrgtj| | kdtj }n| j ||d| d|}t d|j ddd|j d|jd |jd |jd |jd |jd |jd|jd|d|j }|jdkr||jtj }n| dk r.| n|}|j||d|dtd|| j|j|||||||jd| ||j|| f  !            cC s|jdkr%td|jn|jdkrJtd|jn|j|jksk|jtkrtdj|jn|jdkrt|jdkrt nt }n t|j}|jdkr||j |j |_ n||j |_ d|jkod kns7td |jn|jdk rt|jtjr|jtkrtd |jqqt|jd  st|jd rtd|jqnd|j kod knstd|j nt|jtjr|jdkrZ|j dkrNtdttj|j}q|j}qh|jdkrtdttj|j}qh|jdkrtdttj|j}qhtd|jn|jdkr|j}nwt|jtjtjfr|j}nMd|jko5d knr\tt|j|jd}n td||_ t|j!tjtjt"dfstd|j!ndt#t$f}|j%|krtdj||j%ndS(s@Check validity of parameters and raise ValueError if not valid. is.n_estimators must be greater than 0 but was %rgs/learning_rate must be greater than 0 but was %rsLoss '{0:s}' not supported. RiRR'g?s%subsample must be in (0,1] but was %rsinit="%s" is not supportedR+R1sDinit=%r must be valid BaseEstimator and support both fit and predicts&alpha must be in (0.0, 1.0) but was %rRitsqrttlog2sWInvalid value for max_features: %r. Allowed string values are 'auto', 'sqrt' or 'log2'.s'max_features must be in (0, n_features]sCn_iter_no_change should either be None or an integer. %r was passeds,'presort' should be in {}. Got {!r} instead.N(RR'(&RR R_Rtt_SUPPORTED_LOSStLOSS_FUNCTIONSRRtclasses_RRt n_classes_R!RRRR$t isinstanceRt string_typestINIT_ESTIMATORSthasattrRtmaxRR%Rt n_features_RtnumberstIntegraltintegert max_features_RRRReR(R"t loss_classRtallowed_presort((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt _check_paramsst  $ $$       cC s|jdkr$|jj|_n7t|jtjrOt|j|_n |j|_t j |j |jj fdt j |_t j|j fdt j|_|jdkrt j|j dt j|_ndS(sAInitialize model state and allocate model state data structures. R,g?N(RR$RRNtinit_RRRRR%R-RRLtobjectRRuR/RRR(R"((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt _init_state s !cC st|dr-tjd dtj|_nt|drE|`nt|dr]|`nt|dru|`nt|dr|`ndS( s0Clear the state of the gradient boosting model. RiR,RRRt_rngN(ii( RR%R-RRRRRR(R"((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt _clear_states   cC s|j}||jjdkr?td||jdfntj|j||jjf|_tj|j||_|j dkst |drt |drtj|j ||_ qtj |fdtj |_ ndS(s;Add additional ``n_estimators`` entries to all attributes. is(resize with smaller n_estimators %d < %diRR,N(RRR.R R%tresizeRRLRRRRRuR/(R"ttotal_n_estimators((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt _resize_state*s   cC stt|dgdkS(NRi(Rtgetattr(R"((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt_is_initialized>scC st|ddS(sACheck that the estimator is initialized, raising an error if not.RN(R(R"((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt_check_initializedAssPAttribute n_features was deprecated in version 0.19 and will be removed in 0.21.cC s|jS(N(R(R"((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt n_featuresEsc C s |js|jnt||ddddgdt\}}|j\}|_|dkrytj|dtj }nt |dt }t ||||j ||}|jdk rt|||d|jd|j\}}}}}}nd}}}|j|jsf|j|jj||||jj|} d } t|j|_n|j|jjd krtd |j|jjd fn|jjd } t|dtd d dd}|j|} |j|j t krt!|rtd n|j } | dkr;t!| } nd} | rqtj"tj#|dd dtj$} n|j%||| ||j|||| || } | |jjd kr|j| |_|j&| |_&t'|dr|j(| |_(qn| |_)|S(s!Fit the gradient boosting model. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. y : array-like, shape (n_samples,) Target values (strings or integers in classification, real numbers in regression) For classification, labels must correspond to classes. sample_weight : array-like, shape (n_samples,) or None Sample weights. If None, then samples are equally weighted. Splits that would create child nodes with net zero or negative weight are ignored while searching for a split in each node. In the case of classification, splits are also ignored if they would result in any single class carrying a negative weight in either child node. monitor : callable, optional The monitor is called after each iteration with the current iteration, a reference to the estimator and the local variables of ``_fit_stages`` as keyword arguments ``callable(i, self, locals())``. If the callable returns ``True`` the fitting procedure is stopped. The monitor can be used for various things such as computing held-out estimates, early stopping, model introspect, and snapshoting. Returns ------- self : object t accept_sparsetcsrtcsctcooR,twarnRt test_sizeisXn_estimators=%d must be larger or equal to estimators_.shape[0]=%d when warm_start==TruetordertCs0Presorting is not supported for sparse matrices.RRTRN(*RRRRR.RR$R%Rtfloat32RRRt _validate_yRRRRRRRRR+R1RRRRR Rt_decision_functionRRR tasfortranarraytargsorttint32t _fit_stagesRRRt n_estimators_(R"R(R)R*tmonitort n_samplestX_valty_valtsample_weight_valRQRRRtn_stages((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR+Ksb$  *  !         c  C s|jd} |jdk} tj| fdtj}tdt|j| }|j}|jdkr|dk r|jtj |}nd}|j rt |j }|j || nt|rt|nd}t|rt|nd}|jdk r6tj|jtj}|j|}n| }xt| |jD]}| rt| ||}|||||||}n|j|||||||| || }| r||||||||j|<|||||||||j|_s(RR%RuRR/RR:R(R"t total_sumtstaget stage_sumt importances((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytfeature_importances_Rs  cC s4d|_|jjdkr0|jtj}n|S(NitO(RR,tkindRR%R/(R"R)R*((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRgs cC s|j|jdj|dt}|jj\}}tj|jd||f}xft|D]X}xOt|D]A}|j||f}|j|dt |dd||f`. Parameters ---------- loss : {'deviance', 'exponential'}, optional (default='deviance') loss function to be optimized. 'deviance' refers to deviance (= logistic regression) for classification with probabilistic outputs. For loss 'exponential' gradient boosting recovers the AdaBoost algorithm. learning_rate : float, optional (default=0.1) learning rate shrinks the contribution of each tree by `learning_rate`. There is a trade-off between learning_rate and n_estimators. n_estimators : int (default=100) The number of boosting stages to perform. Gradient boosting is fairly robust to over-fitting so a large number usually results in better performance. subsample : float, optional (default=1.0) The fraction of samples to be used for fitting the individual base learners. If smaller than 1.0 this results in Stochastic Gradient Boosting. `subsample` interacts with the parameter `n_estimators`. Choosing `subsample < 1.0` leads to a reduction of variance and an increase in bias. criterion : string, optional (default="friedman_mse") The function to measure the quality of a split. Supported criteria are "friedman_mse" for the mean squared error with improvement score by Friedman, "mse" for mean squared error, and "mae" for the mean absolute error. The default value of "friedman_mse" is generally the best as it can provide a better approximation in some cases. .. versionadded:: 0.18 min_samples_split : int, float, optional (default=2) The minimum number of samples required to split an internal node: - If int, then consider `min_samples_split` as the minimum number. - If float, then `min_samples_split` is a fraction and `ceil(min_samples_split * n_samples)` are the minimum number of samples for each split. .. versionchanged:: 0.18 Added float values for fractions. min_samples_leaf : int, float, optional (default=1) The minimum number of samples required to be at a leaf node. A split point at any depth will only be considered if it leaves at least ``min_samples_leaf`` training samples in each of the left and right branches. This may have the effect of smoothing the model, especially in regression. - If int, then consider `min_samples_leaf` as the minimum number. - If float, then `min_samples_leaf` is a fraction and `ceil(min_samples_leaf * n_samples)` are the minimum number of samples for each node. .. versionchanged:: 0.18 Added float values for fractions. min_weight_fraction_leaf : float, optional (default=0.) The minimum weighted fraction of the sum total of weights (of all the input samples) required to be at a leaf node. Samples have equal weight when sample_weight is not provided. max_depth : integer, optional (default=3) maximum depth of the individual regression estimators. The maximum depth limits the number of nodes in the tree. Tune this parameter for best performance; the best value depends on the interaction of the input variables. min_impurity_decrease : float, optional (default=0.) A node will be split if this split induces a decrease of the impurity greater than or equal to this value. The weighted impurity decrease equation is the following:: N_t / N * (impurity - N_t_R / N_t * right_impurity - N_t_L / N_t * left_impurity) where ``N`` is the total number of samples, ``N_t`` is the number of samples at the current node, ``N_t_L`` is the number of samples in the left child, and ``N_t_R`` is the number of samples in the right child. ``N``, ``N_t``, ``N_t_R`` and ``N_t_L`` all refer to the weighted sum, if ``sample_weight`` is passed. .. versionadded:: 0.19 min_impurity_split : float, (default=1e-7) Threshold for early stopping in tree growth. A node will split if its impurity is above the threshold, otherwise it is a leaf. .. deprecated:: 0.19 ``min_impurity_split`` has been deprecated in favor of ``min_impurity_decrease`` in 0.19. The default value of ``min_impurity_split`` will change from 1e-7 to 0 in 0.23 and it will be removed in 0.25. Use ``min_impurity_decrease`` instead. init : estimator, optional An estimator object that is used to compute the initial predictions. ``init`` has to provide ``fit`` and ``predict``. If None it uses ``loss.init_estimator``. random_state : int, RandomState instance or None, optional (default=None) If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by `np.random`. max_features : int, float, string or None, optional (default=None) The number of features to consider when looking for the best split: - If int, then consider `max_features` features at each split. - If float, then `max_features` is a fraction and `int(max_features * n_features)` features are considered at each split. - If "auto", then `max_features=sqrt(n_features)`. - If "sqrt", then `max_features=sqrt(n_features)`. - If "log2", then `max_features=log2(n_features)`. - If None, then `max_features=n_features`. Choosing `max_features < n_features` leads to a reduction of variance and an increase in bias. Note: the search for a split does not stop until at least one valid partition of the node samples is found, even if it requires to effectively inspect more than ``max_features`` features. verbose : int, default: 0 Enable verbose output. If 1 then it prints progress and performance once in a while (the more trees the lower the frequency). If greater than 1 then it prints progress and performance for every tree. max_leaf_nodes : int or None, optional (default=None) Grow trees with ``max_leaf_nodes`` in best-first fashion. Best nodes are defined as relative reduction in impurity. If None then unlimited number of leaf nodes. warm_start : bool, default: False When set to ``True``, reuse the solution of the previous call to fit and add more estimators to the ensemble, otherwise, just erase the previous solution. See :term:`the Glossary `. presort : bool or 'auto', optional (default='auto') Whether to presort the data to speed up the finding of best splits in fitting. Auto mode by default will use presorting on dense data and default to normal sorting on sparse data. Setting presort to true on sparse data will raise an error. .. versionadded:: 0.17 *presort* parameter. validation_fraction : float, optional, default 0.1 The proportion of training data to set aside as validation set for early stopping. Must be between 0 and 1. Only used if ``n_iter_no_change`` is set to an integer. .. versionadded:: 0.20 n_iter_no_change : int, default None ``n_iter_no_change`` is used to decide if early stopping will be used to terminate training when validation score is not improving. By default it is set to None to disable early stopping. If set to a number, it will set aside ``validation_fraction`` size of the training data as validation and terminate training when validation score is not improving in all of the previous ``n_iter_no_change`` numbers of iterations. .. versionadded:: 0.20 tol : float, optional, default 1e-4 Tolerance for the early stopping. When the loss is not improving by at least tol for ``n_iter_no_change`` iterations (if set to a number), the training stops. .. versionadded:: 0.20 Attributes ---------- n_estimators_ : int The number of estimators as selected by early stopping (if ``n_iter_no_change`` is specified). Otherwise it is set to ``n_estimators``. .. versionadded:: 0.20 feature_importances_ : array, shape (n_features,) The feature importances (the higher, the more important the feature). oob_improvement_ : array, shape (n_estimators,) The improvement in loss (= deviance) on the out-of-bag samples relative to the previous iteration. ``oob_improvement_[0]`` is the improvement in loss of the first stage over the ``init`` estimator. train_score_ : array, shape (n_estimators,) The i-th score ``train_score_[i]`` is the deviance (= loss) of the model at iteration ``i`` on the in-bag sample. If ``subsample == 1`` this is the deviance on the training data. loss_ : LossFunction The concrete ``LossFunction`` object. init_ : estimator The estimator that provides the initial predictions. Set via the ``init`` argument or ``loss.init_estimator``. estimators_ : ndarray of DecisionTreeRegressor,shape (n_estimators, ``loss_.K``) The collection of fitted sub-estimators. ``loss_.K`` is 1 for binary classification, otherwise n_classes. Notes ----- The features are always randomly permuted at each split. Therefore, the best found split may vary, even with the same training data and ``max_features=n_features``, if the improvement of the criterion is identical for several splits enumerated during the search of the best split. To obtain a deterministic behaviour during fitting, ``random_state`` has to be fixed. See also -------- sklearn.tree.DecisionTreeClassifier, RandomForestClassifier AdaBoostClassifier References ---------- J. Friedman, Greedy Function Approximation: A Gradient Boosting Machine, The Annals of Statistics, Vol. 29, No. 5, 2001. J. Friedman, Stochastic Gradient Boosting, 1999 T. Hastie, R. Tibshirani and J. Friedman. Elements of Statistical Learning Ed. 2, Springer, 2009. RRg?idg?t friedman_mseiigiiRg-C6?c+C stt|jd|d|d|d|d|d|d|d| d | d |d |d | d |d|d| d| d|d|d|d|d|dS(NRtR_RRRRRRRRRRRRRRRRRRR(RhR-R#(R"RtR_RRRRRRRRRRRRRRRRRRR((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR#s    cC sxt|tj|dt\|_}tjtj||}|dkrbtd|nt|j|_ |S(Ntreturn_inverseisuy contains %d class after sample_weight trimmed classes with zero weights, while a minimum of 2 classes are required.( RR%RIRRt count_nonzeroRCR RR(R"R)R*tn_trim_classes((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRs   cC sNt|dtdddd}|j|}|jddkrJ|jS|S(sCompute the decision function of ``X``. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Returns ------- score : array, shape (n_samples, n_classes) or (n_samples,) The decision function of the input samples. The order of the classes corresponds to that in the attribute `classes_`. Regression and binary classification produce an array of shape [n_samples]. R,RRRRi(RRRR.Rj(R"R(R((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytdecision_functions  cc s#x|j|D] }|VqWdS(s'Compute decision function of ``X`` for each iteration. This method allows monitoring (i.e. determine error on testing set) after each stage. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Returns ------- score : generator of array, shape (n_samples, k) The decision function of the input samples. The order of the classes corresponds to that in the attribute `classes_`. Regression and binary classification are special cases with ``k == 1``, otherwise ``k==n_classes``. N(R(R"R(tdec((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytstaged_decision_functionscC s7|j|}|jj|}|jj|ddS(sPredict class for X. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Returns ------- y : array, shape (n_samples,) The predicted values. RTi(R2RRRR[(R"R(Rt decisions((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR1scc sGx@|j|D]/}|jj|}|jj|ddVqWdS(s;Predict class at each stage for X. This method allows monitoring (i.e. determine error on testing set) after each stage. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Returns ------- y : generator of array of shape (n_samples,) The predicted value of the input samples. RTiN(RRRRR[(R"R(RR5((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytstaged_predictscC sa|j|}y|jj|SWn7tk r9n$tk r\td|jnXdS(sPredict class probabilities for X. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Raises ------ AttributeError If the ``loss`` does not support probabilities. Returns ------- p : array, shape (n_samples, n_classes) The class probabilities of the input samples. The order of the classes corresponds to that in the attribute `classes_`. s&loss=%r does not support predict_probaN(R2RRRtAttributeErrorRt(R"R(R((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyt predict_probas  cC s|j|}tj|S(sPredict class log-probabilities for X. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Raises ------ AttributeError If the ``loss`` does not support probabilities. Returns ------- p : array, shape (n_samples, n_classes) The class log-probabilities of the input samples. The order of the classes corresponds to that in the attribute `classes_`. (R8R%R<(R"R(R((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytpredict_log_proba-scc smy/x(|j|D]}|jj|VqWWn7tk rEn$tk rhtd|jnXdS(sIPredict class probabilities at each stage for X. This method allows monitoring (i.e. determine error on testing set) after each stage. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Returns ------- y : generator of array of shape (n_samples,) The predicted value of the input samples. s&loss=%r does not support predict_probaN(RRRRR7Rt(R"R(R((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pytstaged_predict_probaEs  (RRN(R2R3R4RR$ReR#RR2R4R1R6R8R9R:(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR-s&         tGradientBoostingRegressorcB szeZdZdZdddddd d d d d ddddd ddedddddZdZdZdZ RS(s0*Gradient Boosting for regression. GB builds an additive model in a forward stage-wise fashion; it allows for the optimization of arbitrary differentiable loss functions. In each stage a regression tree is fit on the negative gradient of the given loss function. Read more in the :ref:`User Guide `. Parameters ---------- loss : {'ls', 'lad', 'huber', 'quantile'}, optional (default='ls') loss function to be optimized. 'ls' refers to least squares regression. 'lad' (least absolute deviation) is a highly robust loss function solely based on order information of the input variables. 'huber' is a combination of the two. 'quantile' allows quantile regression (use `alpha` to specify the quantile). learning_rate : float, optional (default=0.1) learning rate shrinks the contribution of each tree by `learning_rate`. There is a trade-off between learning_rate and n_estimators. n_estimators : int (default=100) The number of boosting stages to perform. Gradient boosting is fairly robust to over-fitting so a large number usually results in better performance. subsample : float, optional (default=1.0) The fraction of samples to be used for fitting the individual base learners. If smaller than 1.0 this results in Stochastic Gradient Boosting. `subsample` interacts with the parameter `n_estimators`. Choosing `subsample < 1.0` leads to a reduction of variance and an increase in bias. criterion : string, optional (default="friedman_mse") The function to measure the quality of a split. Supported criteria are "friedman_mse" for the mean squared error with improvement score by Friedman, "mse" for mean squared error, and "mae" for the mean absolute error. The default value of "friedman_mse" is generally the best as it can provide a better approximation in some cases. .. versionadded:: 0.18 min_samples_split : int, float, optional (default=2) The minimum number of samples required to split an internal node: - If int, then consider `min_samples_split` as the minimum number. - If float, then `min_samples_split` is a fraction and `ceil(min_samples_split * n_samples)` are the minimum number of samples for each split. .. versionchanged:: 0.18 Added float values for fractions. min_samples_leaf : int, float, optional (default=1) The minimum number of samples required to be at a leaf node. A split point at any depth will only be considered if it leaves at least ``min_samples_leaf`` training samples in each of the left and right branches. This may have the effect of smoothing the model, especially in regression. - If int, then consider `min_samples_leaf` as the minimum number. - If float, then `min_samples_leaf` is a fraction and `ceil(min_samples_leaf * n_samples)` are the minimum number of samples for each node. .. versionchanged:: 0.18 Added float values for fractions. min_weight_fraction_leaf : float, optional (default=0.) The minimum weighted fraction of the sum total of weights (of all the input samples) required to be at a leaf node. Samples have equal weight when sample_weight is not provided. max_depth : integer, optional (default=3) maximum depth of the individual regression estimators. The maximum depth limits the number of nodes in the tree. Tune this parameter for best performance; the best value depends on the interaction of the input variables. min_impurity_decrease : float, optional (default=0.) A node will be split if this split induces a decrease of the impurity greater than or equal to this value. The weighted impurity decrease equation is the following:: N_t / N * (impurity - N_t_R / N_t * right_impurity - N_t_L / N_t * left_impurity) where ``N`` is the total number of samples, ``N_t`` is the number of samples at the current node, ``N_t_L`` is the number of samples in the left child, and ``N_t_R`` is the number of samples in the right child. ``N``, ``N_t``, ``N_t_R`` and ``N_t_L`` all refer to the weighted sum, if ``sample_weight`` is passed. .. versionadded:: 0.19 min_impurity_split : float, (default=1e-7) Threshold for early stopping in tree growth. A node will split if its impurity is above the threshold, otherwise it is a leaf. .. deprecated:: 0.19 ``min_impurity_split`` has been deprecated in favor of ``min_impurity_decrease`` in 0.19. The default value of ``min_impurity_split`` will change from 1e-7 to 0 in 0.23 and it will be removed in 0.25. Use ``min_impurity_decrease`` instead. init : estimator, optional (default=None) An estimator object that is used to compute the initial predictions. ``init`` has to provide ``fit`` and ``predict``. If None it uses ``loss.init_estimator``. random_state : int, RandomState instance or None, optional (default=None) If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by `np.random`. max_features : int, float, string or None, optional (default=None) The number of features to consider when looking for the best split: - If int, then consider `max_features` features at each split. - If float, then `max_features` is a fraction and `int(max_features * n_features)` features are considered at each split. - If "auto", then `max_features=n_features`. - If "sqrt", then `max_features=sqrt(n_features)`. - If "log2", then `max_features=log2(n_features)`. - If None, then `max_features=n_features`. Choosing `max_features < n_features` leads to a reduction of variance and an increase in bias. Note: the search for a split does not stop until at least one valid partition of the node samples is found, even if it requires to effectively inspect more than ``max_features`` features. alpha : float (default=0.9) The alpha-quantile of the huber loss function and the quantile loss function. Only if ``loss='huber'`` or ``loss='quantile'``. verbose : int, default: 0 Enable verbose output. If 1 then it prints progress and performance once in a while (the more trees the lower the frequency). If greater than 1 then it prints progress and performance for every tree. max_leaf_nodes : int or None, optional (default=None) Grow trees with ``max_leaf_nodes`` in best-first fashion. Best nodes are defined as relative reduction in impurity. If None then unlimited number of leaf nodes. warm_start : bool, default: False When set to ``True``, reuse the solution of the previous call to fit and add more estimators to the ensemble, otherwise, just erase the previous solution. See :term:`the Glossary `. presort : bool or 'auto', optional (default='auto') Whether to presort the data to speed up the finding of best splits in fitting. Auto mode by default will use presorting on dense data and default to normal sorting on sparse data. Setting presort to true on sparse data will raise an error. .. versionadded:: 0.17 optional parameter *presort*. validation_fraction : float, optional, default 0.1 The proportion of training data to set aside as validation set for early stopping. Must be between 0 and 1. Only used if ``n_iter_no_change`` is set to an integer. .. versionadded:: 0.20 n_iter_no_change : int, default None ``n_iter_no_change`` is used to decide if early stopping will be used to terminate training when validation score is not improving. By default it is set to None to disable early stopping. If set to a number, it will set aside ``validation_fraction`` size of the training data as validation and terminate training when validation score is not improving in all of the previous ``n_iter_no_change`` numbers of iterations. .. versionadded:: 0.20 tol : float, optional, default 1e-4 Tolerance for the early stopping. When the loss is not improving by at least tol for ``n_iter_no_change`` iterations (if set to a number), the training stops. .. versionadded:: 0.20 Attributes ---------- feature_importances_ : array, shape (n_features,) The feature importances (the higher, the more important the feature). oob_improvement_ : array, shape (n_estimators,) The improvement in loss (= deviance) on the out-of-bag samples relative to the previous iteration. ``oob_improvement_[0]`` is the improvement in loss of the first stage over the ``init`` estimator. train_score_ : array, shape (n_estimators,) The i-th score ``train_score_[i]`` is the deviance (= loss) of the model at iteration ``i`` on the in-bag sample. If ``subsample == 1`` this is the deviance on the training data. loss_ : LossFunction The concrete ``LossFunction`` object. init_ : estimator The estimator that provides the initial predictions. Set via the ``init`` argument or ``loss.init_estimator``. estimators_ : array of DecisionTreeRegressor, shape (n_estimators, 1) The collection of fitted sub-estimators. Notes ----- The features are always randomly permuted at each split. Therefore, the best found split may vary, even with the same training data and ``max_features=n_features``, if the improvement of the criterion is identical for several splits enumerated during the search of the best split. To obtain a deterministic behaviour during fitting, ``random_state`` has to be fixed. See also -------- DecisionTreeRegressor, RandomForestRegressor References ---------- J. Friedman, Greedy Function Approximation: A Gradient Boosting Machine, The Annals of Statistics, Vol. 29, No. 5, 2001. J. Friedman, Stochastic Gradient Boosting, 1999 T. Hastie, R. Tibshirani and J. Friedman. Elements of Statistical Learning Ed. 2, Springer, 2009. RRRR'g?idg?R.iigig?iRg-C6?c-C stt|jd|d|d|d|d|d|d|d| d | d |d |d | d | d| d|d|d|d|d|d|d|d|dS(NRtR_RRRRRRRRRRRRR!RRRRRRR(RhR;R#(R"RtR_RRRRRRRRRRRRR!RRRRRRR((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR#W s    cC s1t|dtdddd}|j|jS(sPredict regression target for X. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Returns ------- y : array, shape (n_samples,) The predicted values. R,RRRR(RRRRj(R"R(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR1n scc s)x"|j|D]}|jVqWdS(sGPredict regression target at each stage for X. This method allows monitoring (i.e. determine error on testing set) after each stage. Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, it will be converted to ``dtype=np.float32`` and if a sparse matrix is provided to a sparse ``csr_matrix``. Returns ------- y : generator of array of shape (n_samples,) The predicted value of the input samples. N(RRj(R"R(R)((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR6 scC s?tt|j|}|j|jd|jjd}|S(sApply trees in the ensemble to X, return leaf indices. .. versionadded:: 0.17 Parameters ---------- X : {array-like, sparse matrix}, shape (n_samples, n_features) The input samples. Internally, its dtype will be converted to ``dtype=np.float32``. If a sparse matrix is provided, it will be converted to a sparse ``csr_matrix``. Returns ------- X_leaves : array-like, shape (n_samples, n_estimators) For each datapoint x in X and for each tree in the ensemble, return the index of the leaf x ends up in each estimator. i(RhR;RUtreshapeR.R(R"R(R*((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyRU s#(RRRR'N( R2R3R4RR$ReR#R1R6RU(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyR;as       (LR4t __future__RRtabcRRtbaseRRRt externalsRt_gradient_boostingRR R RtnumpyR%t scipy.sparseR R R t scipy.specialRRtmodel_selectionRt tree.treeRt tree._treeRRtutilsRRRRRRt utils.fixesRt utils.statsRtutils.validationRtutils.multiclassRt exceptionsRRRR5R9R@RARFtwith_metaclassRKRgRiRkRoRzR}RRRR$RRRRR-R;(((sAlib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyts~  5)2+."k"S8gM"[_W   K"4