bcg5ddlmZddlZddlmZddlmZddlmZmZddl m Z m Z dd l m Z dd lmZdd lmZdd lmZdd lmZmZGddeZGddeZy))print_functionN)copy)Enum) CatBoostErrorCatBoost)EvaluationResultsMetricEvaluationResult)FoldModelsHandler)_SimpleStreamingFileReader) _Splitter) ExecutionCase) LabelMode FactorUtilsc eZdZdZdZdZdZdZy)EvalTypez Type of feature evaluation: All: All factors presented SeqRem: Each factor while other presented SeqAdd: Each factor while other removed SeqAddAndAll: SeqAdd + All AllSeqRemSeqAdd SeqAddAndAllN)__name__ __module__ __qualname____doc__rrrr\/var/www/html/bid-api/venv/lib/python3.12/site-packages/catboost/eval/catboost_evaluation.pyrrs C F F!LrrceZdZ d dZdZedZedZdZd dZ de jddde jfd Z d d Zy) CatboostEvaluationNc ddl} | j|_| jj |j||_|&| jj |j|nd|_||_||_||_ | |_ | |_ | |_ t| |_||_|t||_n||_|ddl}|j%|_y||_y)ai Args: :param path_to_dataset: (str) Path to the dataset to be used for evaluation. :param fold_size: (int) Size of the folds in cross-validation. :param fold_count: (int) Number of times we get a new fold, learn a model and check results as if there wouldn't be any offset. If there'are some offset it means that the real count of folds will be smaller. :param column_description: (str) Path to the file where column description is placed. :param fold_offset: (int) Number of folds we skip before begin to make cross-validation. :param group_column: (int) GroupId column index in the dataset file. 'None' value means absence of grouping information in the dataset (it's the default). :param working_dir: Working dir for temporary files :param remove_models: (bool) Set true if you want models to be removed after applying them. :param delimiter: (str) Field delimiter used in dataset files. :param has_header: (bool) Set true if you want to skip first line in dataset files. :param partition_random_seed: (int) The seed for random value generator used for getting permutations for cross-validation. :param min_fold_count: (int) Minimun amount of folds dataset can be cut to. rN)os.pathgetcwd _current_dirpathjoin_path_to_dataset_column_description _fold_offset _fold_count _fold_size _delimiter _has_header_seedint_min_fold_count_remove_models_group_feature_numtempfilemkdtemp _working_dir)selfpath_to_dataset fold_size fold_countcolumn_description fold_offset group_column working_dir remove_models delimiter has_headerpartition_random_seedmin_fold_countosr3s r__init__zCatboostEvaluation.__init__ s@ %BIIK " T->-> PGYGe$&77<<0A0A0B$Dko  '%##%* ">2+  #&),&7D #&2D #    ( 0 0 2D  +D rcltj}tj|j|SN)rCr#chdirr5)r6currents r__go_to_working_dirz&CatboostEvaluation.__go_to_working_dirZs$))+ ""#rc t|dk(r tdt}t|t|t j |||}|t jk(s!|t jk(st|dk(r7|jt|gt j |g|||fS|t jk(r@|D]7}|jt||gt j ||g|9||fS|t jk(s|t jk(rc|D]Z}t|}|j||jt|t j |||t|\||fS|t jk7rtd|j||fS)Nrz*Provide at least one feature to evaluation) label_mode)ignored_featureslabelr )rMrLzDon't support {} mode.)lenrlistrr create_labelrrrappendrrrremoveAttributeErrorvalue)paramsfeatures_to_eval eval_typerK test_cases baseline_case feature_num cur_featuress r_create_eval_feature_casesz-CatboostEvaluation._create_eval_feature_cases_s  A % LM MV %f7;)sepr@group_feature_num)seedmin_folds_count)splitterr8 folds_countr;)setrNr&_CatboostEvaluation__go_to_working_dirr*r)rSformatr r1r r'r,r-r2rr(r.r0proceedr+rhrCrG) r6rmrnrpevaluation_step cases_set current_wd error_msghandlerreaderruresults r_calculate_result_metricsz,CatboostEvaluation._calculate_result_metricssYJ y>SZ ' ;e CD D--/  !4#4#44v $Y%5%5d6G6G6:6F6F&HII(e072A5A6:6I6I KG 00E0E48OO;?;K;KBFBYBY[F!!%!9!9&*jj151E1EGH __h/3151A1A151B1B%DF 226: HHZ BHHZ s D E##E:c t|}|g}t|tr|n|g}t|tr|j } n t |} |*d| vr5| d|k7r-t dj| d|d| vr t d| |dk7r|| d<|d}||| d<n| d}| ddk(r t d|j| ||| \} } ||vr|j||j| | ||| S) a Evaluate features. Args: learn_config: dict with params or instance of CatBoost. In second case instance params will be used features_to_eval: list of indices of features to evaluate loss_function: one of CatBoost loss functions, get it from learn_config if not specified eval_type: Type of feature evaluate (All, SeqAdd, SeqRem) eval_metrics: Additional metrics to calculate thread_count: thread_count to use. If not none will override learn_config values Returns ------- result : Instance of EvaluationResult class loss_functionzVLoss function in params {} should be equal to feature evaluation objective function {}zDProvide loss function in params or as option to eval_features methodrkrpr PairLogitz$Pair classification is not supported)rWrK)rY compare_cases eval_metricsrpro) rw isinstancerOr get_paramsr^rryr\rQ eval_cases) r6 learn_configrVrrWrrprorKrUrYrXs r eval_featuresz CatboostEvaluation.eval_featuress]*/0  L'1,'E|L> lH -!,,.F,'F  $&(VO-D -U#%228&9PR_2`bbf,#$jkk  # (:%1F> "  I  $&3F? #"?3M / "k 1 FG G$($C$CFDTNWOY%D%[! z  ,    .]-7,8,8)2 4 4rct|ts|g}|g}||z }|D]}|j||j||||}|j ||S)aMore flexible evaluation of any cases. Args: baseline_case: Execution case used for baseline compare_cases: List of cases to compare eval_metrics: Metrics to calculate thread_count: thread_count to use. Will override one in cases Returns ------- result : Instance of EvaluationResult class )rpr{)rrO_set_thread_countrset_baseline_case) r6rYrrrprormrb metric_results rrzCatboostEvaluation.eval_cases sw -.*OM D  " "< 066u7CDPGP7R  '' 6r)rNNT Frr )rkr )rrrrDrx staticmethodr\rhrjrrrr AddFeaturerrrrrr r s"!#!'( !8,t ')')R * *!+!`%) (#'#% $!*!5!5?4J!# rr ) __future__rrCrenumrrrrer r _fold_models_handlerr _readersr _splitterrexecution_caser factor_utilsrrrobjectr rrrrs@% &H30 )0 "t "IIr