upgini 1.2.16a1__tar.gz → 1.2.16a3654.dev2__tar.gz

{upgini-1.2.16a1 → upgini-1.2.16a3654.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: upgini
-Version: 1.2.16a1
+Version: 1.2.16a3654.dev2
 Summary: Intelligent data search & enrichment for Machine Learning
 Project-URL: Bug Reports, https://github.com/upgini/upgini/issues
 Project-URL: Homepage, https://upgini.com/

upgini-1.2.16a3654.dev2/src/upgini/__about__.py

@@ -0,0 +1 @@
+__version__ = "1.2.16a3654.dev2"

{upgini-1.2.16a1 → upgini-1.2.16a3654.dev2}/src/upgini/features_enricher.py

@@ -165,6 +165,7 @@ class FeaturesEnricher(TransformerMixin):
     RANDOM_STATE = 42
     CALCULATE_METRICS_THRESHOLD = 50_000_000
     CALCULATE_METRICS_MIN_THRESHOLD = 500
+    TEXT_FEATURES_THRESHOLD = 5_000
     GENERATE_FEATURES_LIMIT = 10
     EMPTY_FEATURES_INFO = pd.DataFrame(
         columns=[
@@ -871,13 +872,6 @@ class FeaturesEnricher(TransformerMixin):
                 else None
             )
 
-            if self.X is None:
-                self.X = X
-            if self.y is None:
-                self.y = y
-            if self.eval_set is None:
-                self.eval_set = effective_eval_set
-
             try:
                 self.__log_debug_information(
                     validated_X,
@@ -945,14 +939,16 @@ class FeaturesEnricher(TransformerMixin):
 
                 gc.collect()
 
-                if fitting_X.shape[1] == 0 and fitting_enriched_X.shape[1] == 0:
-                    print(self.bundle.get("metrics_no_important_free_features"))
-                    self.logger.warning("No client or free relevant ADS features found to calculate metrics")
-                    self.warning_counter.increment()
-                    return None
+                text_features = self.generate_features if fitting_X.shape[0] >= self.TEXT_FEATURES_THRESHOLD else []
 
                 print(self.bundle.get("metrics_start"))
                 with Spinner():
+                    if fitting_X.shape[1] == 0 and fitting_enriched_X.shape[1] == 0:
+                        print(self.bundle.get("metrics_no_important_free_features"))
+                        self.logger.warning("No client or free relevant ADS features found to calculate metrics")
+                        self.warning_counter.increment()
+                        return None
+
                     self._check_train_and_eval_target_distribution(y_sorted, fitting_eval_set_dict)
 
                     has_date = SearchKey.find_key(search_keys, [SearchKey.DATE, SearchKey.DATETIME]) is not None
@@ -966,7 +962,7 @@ class FeaturesEnricher(TransformerMixin):
                         fitting_enriched_X,
                         scoring,
                         groups=groups,
-                        text_features=self.generate_features,
+                        text_features=text_features,
                         has_date=has_date,
                     )
                     metric = wrapper.metric_name
@@ -993,10 +989,10 @@ class FeaturesEnricher(TransformerMixin):
                             cat_features,
                             add_params=custom_loss_add_params,
                             groups=groups,
-                            text_features=self.generate_features,
+                            text_features=text_features,
                             has_date=has_date,
                         )
-                        etalon_metric, _ = baseline_estimator.cross_val_predict(
+                        etalon_metric = baseline_estimator.cross_val_predict(
                             fitting_X, y_sorted, self.baseline_score_column
                         )
                         if etalon_metric is None:
@@ -1027,14 +1023,10 @@ class FeaturesEnricher(TransformerMixin):
                             cat_features,
                             add_params=custom_loss_add_params,
                             groups=groups,
-                            text_features=self.generate_features,
+                            text_features=text_features,
                             has_date=has_date,
                         )
-                        enriched_metric, enriched_shaps = enriched_estimator.cross_val_predict(
-                            fitting_enriched_X, enriched_y_sorted
-                        )
-
-                        print(f"Calculated enriched shaps: {enriched_shaps}")
+                        enriched_metric = enriched_estimator.cross_val_predict(fitting_enriched_X, enriched_y_sorted)
                         if enriched_metric is None:
                             self.logger.warning(
                                 f"Enriched {metric} on train combined features is None (maybe all features was removed)"
@@ -1167,6 +1159,13 @@ class FeaturesEnricher(TransformerMixin):
                     elif uplift_col in metrics_df.columns and (metrics_df[uplift_col] < 0).any():
                         self.logger.warning("Uplift is negative")
 
+                    if self.X is None:
+                        self.X = X
+                    if self.y is None:
+                        self.y = y
+                    if self.eval_set is None:
+                        self.eval_set = effective_eval_set
+
                     return metrics_df
             except Exception as e:
                 error_message = "Failed to calculate metrics" + (
@@ -1191,18 +1190,6 @@ class FeaturesEnricher(TransformerMixin):
             finally:
                 self.logger.info(f"Calculating metrics elapsed time: {time.time() - start_time}")
 
-    def _update_shap_values(self, new_shaps: Dict[str, float]):
-        feature_name_header = self.bundle.get("features_info_name")
-        shap_value_header = self.bundle.get("features_info_shap")
-
-        def update_shap(row):
-            return new_shaps.get(row[feature_name_header], row[shap_value_header])
-
-        self.features_info[shap_value_header] = self.features_info.apply(update_shap, axis=1)
-        self.features_info.sort_values(by=shap_value_header, ascending=False, inplace=True)
-
-        # TODO redraw
-
     def _check_train_and_eval_target_distribution(self, y, eval_set_dict):
         uneven_distribution = False
         for eval_set in eval_set_dict.values():
@@ -1594,9 +1581,7 @@ class FeaturesEnricher(TransformerMixin):
             generated_features.extend(generator.generated_features)
 
         normalizer = Normalizer(self.bundle, self.logger, self.warning_counter)
-        df, search_keys, generated_features = normalizer.normalize(
-            df, search_keys, generated_features
-        )
+        df, search_keys, generated_features = normalizer.normalize(df, search_keys, generated_features)
         columns_renaming = normalizer.columns_renaming
 
         df = clean_full_duplicates(df, logger=self.logger, silent=True, bundle=self.bundle)
@@ -2036,9 +2021,7 @@ class FeaturesEnricher(TransformerMixin):
                 generated_features.extend(generator.generated_features)
 
             normalizer = Normalizer(self.bundle, self.logger, self.warning_counter, silent_mode)
-            df, search_keys, generated_features = normalizer.normalize(
-                df, search_keys, generated_features
-            )
+            df, search_keys, generated_features = normalizer.normalize(df, search_keys, generated_features)
             columns_renaming = normalizer.columns_renaming
 
             # Don't pass all features in backend on transform

{upgini-1.2.16a1 → upgini-1.2.16a3654.dev2}/src/upgini/metrics.py

@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-from collections import defaultdict
 import inspect
 import logging
 import re
@@ -10,7 +9,7 @@ from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 import catboost
 import numpy as np
 import pandas as pd
-from catboost import CatBoost, CatBoostClassifier, CatBoostRegressor, Pool
+from catboost import CatBoostClassifier, CatBoostRegressor
 from numpy import log1p
 from pandas.api.types import is_numeric_dtype
 from sklearn.metrics import check_scoring, get_scorer, make_scorer, roc_auc_score
@@ -64,7 +63,7 @@ CATBOOST_BINARY_PARAMS = {
     "verbose": False,
     "random_state": DEFAULT_RANDOM_STATE,
     "allow_writing_files": False,
-    "auto_class_weights": "Balanced",
+    "auto_class_weights": "SqrtBalanced",
 }
 
 CATBOOST_MULTICLASS_PARAMS = {
@@ -82,7 +81,7 @@ CATBOOST_MULTICLASS_PARAMS = {
     "verbose": False,
     "random_state": DEFAULT_RANDOM_STATE,
     "allow_writing_files": False,
-    "auto_class_weights": "Balanced",
+    "auto_class_weights": "SqrtBalanced",
 }
 
 LIGHTGBM_PARAMS = {
@@ -289,12 +288,9 @@ class EstimatorWrapper:
         x, y, _ = self._prepare_data(x, y)
         return x, y, {}
 
-    def calculate_shap(self, x: pd.DataFrame, y: pd.Series, estimator) -> Optional[Dict[str, float]]:
-        return None
-
     def cross_val_predict(
         self, x: pd.DataFrame, y: np.ndarray, baseline_score_column: Optional[Any] = None
-    ) -> Tuple[Optional[float], Optional[Dict[str, float]]]:
+    ) -> Optional[float]:
         x, y, groups, fit_params = self._prepare_to_fit(x, y)
 
         if x.shape[1] == 0:
@@ -302,7 +298,6 @@ class EstimatorWrapper:
 
         scorer = check_scoring(self.estimator, scoring=self.scorer)
 
-        shap_values_all_folds = defaultdict(list)
         if baseline_score_column is not None and self.metric_name == "GINI":
             self.logger.info("Calculate baseline GINI on passed baseline_score_column and target")
             metric = roc_auc_score(y, x[baseline_score_column])
@@ -324,29 +319,7 @@ class EstimatorWrapper:
             self.check_fold_metrics(metrics_by_fold)
 
             metric = np.mean(metrics_by_fold) * self.multiplier
-
-            splits = self.cv.split(x, y, groups)
-
-            for estimator, split in zip(self.cv_estimators, splits):
-                _, validation_idx = split
-                cv_x = x.iloc[validation_idx]
-                cv_y = y[validation_idx]
-                shaps = self.calculate_shap(cv_x, cv_y, estimator)
-                if shaps is not None:
-                    for feature, shap_value in shaps.items():
-                        # shap_values_all_folds[feature] = shap_values_all_folds.get(feature, []) + shap_value.tolist()
-                        shap_values_all_folds[feature].extend(shap_value.tolist())
-
-        if shap_values_all_folds:
-            average_shap_values = {
-                feature: np.mean(shaps)
-                for feature, shaps
-                in shap_values_all_folds.items()
-            }
-        else:
-            average_shap_values = None
-
-        return self.post_process_metric(metric), average_shap_values
+        return self.post_process_metric(metric)
 
     def check_fold_metrics(self, metrics_by_fold: List[float]):
         first_metric_sign = 1 if metrics_by_fold[0] >= 0 else -1
@@ -575,7 +548,7 @@ class CatBoostWrapper(EstimatorWrapper):
 
     def cross_val_predict(
         self, x: pd.DataFrame, y: np.ndarray, baseline_score_column: Optional[Any] = None
-    ) -> Tuple[Optional[float], Optional[Dict[str, float]]]:
+    ) -> Optional[float]:
         try:
             return super().cross_val_predict(x, y, baseline_score_column)
         except Exception as e:
@@ -600,29 +573,6 @@ class CatBoostWrapper(EstimatorWrapper):
             else:
                 raise e
 
-    def calculate_shap(self, x: pd.DataFrame, y: pd.Series, estimator: CatBoost) -> Optional[Dict[str, float]]:
-        try:
-            # Create Pool for fold data, if need (for example, when categorical features are present)
-            fold_pool = Pool(x, y, cat_features=self.cat_features)
-
-            # Get SHAP values of current estimator
-            shap_values_fold = estimator.get_feature_importance(data=fold_pool, type="ShapValues")
-
-            # Remove last columns (base value) and flatten
-            if self.target_type == ModelTaskType.MULTICLASS:
-                all_shaps = shap_values_fold[:, :, :-1]
-                all_shaps = [all_shaps[:, :, k].flatten() for k in range(all_shaps.shape[2])]
-            else:
-                all_shaps = shap_values_fold[:, :-1]
-                all_shaps = [all_shaps[:, k].flatten() for k in range(all_shaps.shape[1])]
-
-            all_shaps = np.abs(all_shaps)
-
-            return dict(zip(estimator.feature_names_, all_shaps))
-
-        except Exception:
-            return None
-
 
 class LightGBMWrapper(EstimatorWrapper):
     def __init__(

{upgini-1.2.16a1 → upgini-1.2.16a3654.dev2}/src/upgini/normalizer/normalize_utils.py

@@ -49,10 +49,7 @@ class Normalizer:
         self.generated_features = []
 
     def normalize(
-        self,
-        df: pd.DataFrame,
-        search_keys: Dict[str, SearchKey],
-        generated_features: List[str],
+        self, df: pd.DataFrame, search_keys: Dict[str, SearchKey], generated_features: List[str]
     ) -> Tuple[pd.DataFrame, Dict[str, SearchKey], List[str]]:
         self.search_keys = search_keys.copy()
         self.generated_features = generated_features.copy()

{upgini-1.2.16a1 → upgini-1.2.16a3654.dev2}/src/upgini/utils/display_utils.py

@@ -72,7 +72,7 @@ def make_table(df: pd.DataFrame, wrap_long_string=None) -> str:
     )
 
 
-def display_html_dataframe(df: pd.DataFrame, internal_df: pd.DataFrame, header: str, display_id: str):
+def display_html_dataframe(df: pd.DataFrame, internal_df: pd.DataFrame, header: str):
     if not ipython_available():
         print(header)
         print(internal_df)
@@ -133,7 +133,7 @@ def display_html_dataframe(df: pd.DataFrame, internal_df: pd.DataFrame, header:
             {table_html}
         </div>
         """
-    return display(HTML(result_html))
+    display(HTML(result_html))
 
 
 def make_html_report(

{upgini-1.2.16a1 → upgini-1.2.16a3654.dev2}/src/upgini/utils/features_validator.py

@@ -58,10 +58,6 @@ class FeaturesValidator:
 
         columns_renaming = columns_renaming or {}
 
-        if features_for_generate:
-            empty_or_constant_features = [
-                f for f in empty_or_constant_features if columns_renaming.get(f, f) not in features_for_generate
-            ]
         if empty_or_constant_features:
             msg = bundle.get("empty_or_contant_features").format(
                 [columns_renaming.get(f, f) for f in empty_or_constant_features]

upgini-1.2.16a1/src/upgini/__about__.py

@@ -1 +0,0 @@
-__version__ = "1.2.16a1"