upgini 1.2.71a3832.dev12__py3-none-any.whl → 1.2.72__py3-none-any.whl

upgini/__about__.py

@@ -1 +1 @@
-__version__ = "1.2.71a3832.dev12"
+__version__ = "1.2.72"

upgini/features_enricher.py

@@ -3250,8 +3250,7 @@ if response.status_code == 200:
     def _validate_eval_set_pair(self, X: pd.DataFrame, eval_pair: Tuple) -> Tuple[pd.DataFrame, pd.Series]:
         if len(eval_pair) != 2:
             raise ValidationError(self.bundle.get("eval_set_invalid_tuple_size").format(len(eval_pair)))
-        eval_X = eval_pair[0]
-        eval_y = eval_pair[1]
+        eval_X, eval_y = eval_pair
 
         if _num_samples(eval_X) == 0:
             raise ValidationError(self.bundle.get("eval_x_is_empty"))
@@ -3872,15 +3871,23 @@ if response.status_code == 200:
 
         original_shaps = {original_names_dict.get(fm.name, fm.name): fm.shap_value for fm in features_meta}
 
-        if updated_shaps is not None:
-            for fm in features_meta:
-                fm.shap_value = updated_shaps.get(fm.name, 0.0)
-
-        features_meta.sort(key=lambda m: (-m.shap_value, m.name))
         for feature_meta in features_meta:
             if feature_meta.name in original_names_dict.keys():
                 feature_meta.name = original_names_dict[feature_meta.name]
 
+            if updated_shaps is not None:
+                updating_shap = updated_shaps.get(feature_meta.name)
+                if updating_shap is None:
+                    self.logger.warning(
+                        f"WARNING: Shap value for feature {feature_meta.name} not found and will be set to 0.0"
+                    )
+                    updating_shap = 0.0
+                feature_meta.shap_value = updating_shap
+
+        features_meta.sort(key=lambda m: (-m.shap_value, m.name))
+
+        for feature_meta in features_meta:
+
             is_client_feature = feature_meta.name in df.columns
 
             # TODO make a decision about selected features based on special flag from mlb
@@ -3892,7 +3899,7 @@ if response.status_code == 200:
             # Use only important features
             if (
                 # feature_meta.name in self.fit_generated_features or
-                feature_meta.name == COUNTRY
+                feature_meta.name == COUNTRY  # constant synthetic column
                 # In select_features mode we select also from etalon features and need to show them
                 or (not self.fit_select_features and is_client_feature)
             ):

upgini/metrics.py

@@ -8,16 +8,18 @@ from copy import deepcopy
 from dataclasses import dataclass
 from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
+import lightgbm as lgb
 import numpy as np
 import pandas as pd
 from lightgbm import LGBMClassifier, LGBMRegressor
-import lightgbm as lgb
 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
+from sklearn.preprocessing import OrdinalEncoder
 
 from upgini.utils.features_validator import FeaturesValidator
 from upgini.utils.sklearn_ext import cross_validate
+from upgini.utils.blocked_time_series import BlockedTimeSeriesSplit
 
 try:
     from sklearn.metrics import get_scorer_names
@@ -29,7 +31,7 @@ except ImportError:
     available_scorers = SCORERS
 from sklearn.metrics import mean_squared_error
 from sklearn.metrics._regression import _check_reg_targets, check_consistent_length
-from sklearn.model_selection import BaseCrossValidator
+from sklearn.model_selection import BaseCrossValidator, TimeSeriesSplit
 
 from upgini.errors import ValidationError
 from upgini.metadata import ModelTaskType
@@ -83,22 +85,6 @@ CATBOOST_MULTICLASS_PARAMS = {
     "auto_class_weights": "Balanced",
 }
 
-LIGHTGBM_PARAMS = {
-    "random_state": DEFAULT_RANDOM_STATE,
-    # "num_leaves": 16,
-    # "n_estimators": 150,
-    # "min_child_weight": 1,
-    "max_depth": 4,
-    "max_cat_threshold": 80,
-    "min_data_per_group": 25,
-    "num_boost_round": 150,
-    "cat_l2": 10,
-    "cat_smooth": 12,
-    "learning_rate": 0.05,
-    "feature_fraction": 1.0,
-    "min_sum_hessian_in_leaf": 0.01,
-}
-
 LIGHTGBM_REGRESSION_PARAMS = {
     "random_state": DEFAULT_RANDOM_STATE,
     "deterministic": True,
@@ -125,9 +111,9 @@ LIGHTGBM_MULTICLASS_PARAMS = {
     "max_cat_threshold": 80,
     "min_data_per_group": 20,
     "cat_smooth": 18,
-    "cat_l2" : 8,
+    "cat_l2": 8,
     "objective": "multiclass",
-    "class_weight": "balanced",
+    # "class_weight": "balanced",
     "use_quantized_grad": "true",
     "num_grad_quant_bins": "8",
     "stochastic_rounding": "true",
@@ -141,12 +127,12 @@ LIGHTGBM_BINARY_PARAMS = {
     "max_depth": 5,
     "learning_rate": 0.05,
     "objective": "binary",
-    "class_weight": "balanced",
+    # "class_weight": "balanced",
     "deterministic": True,
     "max_cat_threshold": 80,
     "min_data_per_group": 20,
     "cat_smooth": 18,
-    "cat_l2" : 8,
+    "cat_l2": 8,
     "verbosity": -1,
 }
 
@@ -505,7 +491,8 @@ class EstimatorWrapper:
                 params = _get_add_params(params, add_params)
                 estimator = LightGBMWrapper(LGBMClassifier(**params), **kwargs)
             elif target_type == ModelTaskType.REGRESSION:
-                params = _get_add_params(params, LIGHTGBM_REGRESSION_PARAMS)
+                if not isinstance(cv, TimeSeriesSplit) and not isinstance(cv, BlockedTimeSeriesSplit):
+                    params = _get_add_params(params, LIGHTGBM_REGRESSION_PARAMS)
                 params = _get_add_params(params, add_params)
                 estimator = LightGBMWrapper(LGBMRegressor(**params), **kwargs)
             else:
@@ -754,6 +741,7 @@ class LightGBMWrapper(EstimatorWrapper):
             logger=logger,
         )
         self.cat_features = None
+        self.cat_encoder = None
         self.n_classes = None
 
     def _prepare_to_fit(self, x: pd.DataFrame, y: pd.Series) -> Tuple[pd.DataFrame, pd.Series, np.ndarray, dict]:
@@ -764,10 +752,13 @@ class LightGBMWrapper(EstimatorWrapper):
             params["callbacks"] = [lgb.early_stopping(stopping_rounds=LIGHTGBM_EARLY_STOPPING_ROUNDS, verbose=False)]
         self.cat_features = _get_cat_features(x)
         if self.cat_features:
-            params["categorical_feature"] = self.cat_features
-        x = fill_na_cat_features(x, self.cat_features)
-        for feature in self.cat_features:
-            x[feature] = x[feature].astype("category").cat.codes
+            x = fill_na_cat_features(x, self.cat_features)
+            encoder = OrdinalEncoder(handle_unknown="use_encoded_value", unknown_value=-1)
+            encoded = pd.DataFrame(
+                encoder.fit_transform(x[self.cat_features]), columns=self.cat_features, dtype="category"
+            )
+            x[self.cat_features] = encoded
+            self.cat_encoder = encoder
         if not is_numeric_dtype(y_numpy):
             y_numpy = correct_string_target(y_numpy)
 
@@ -777,8 +768,10 @@ class LightGBMWrapper(EstimatorWrapper):
         x, y_numpy, params = super()._prepare_to_calculate(x, y)
         if self.cat_features is not None:
             x = fill_na_cat_features(x, self.cat_features)
-            for feature in self.cat_features:
-                x[feature] = x[feature].astype("category").cat.codes
+            if self.cat_encoder is not None:
+                x[self.cat_features] = pd.DataFrame(
+                    self.cat_encoder.transform(x[self.cat_features]), columns=self.cat_features, dtype="category"
+                )
         if not is_numeric_dtype(y):
             y_numpy = correct_string_target(y_numpy)
         return x, y_numpy, params

upgini/utils/target_utils.py

@@ -204,7 +204,7 @@ def balance_undersample(
 def balance_undersample_forced(
     df: pd.DataFrame,
     target_column: str,
-    id_columns: List[str],
+    id_columns: Optional[List[str]],
     date_column: str,
     task_type: ModelTaskType,
     cv_type: Optional[CVType],
@@ -287,7 +287,7 @@ DEFAULT_TIME_UNIT_THRESHOLD = pd.Timedelta(weeks=4)
 
 def balance_undersample_time_series_trunc(
     df: pd.DataFrame,
-    id_columns: List[str],
+    id_columns: Optional[List[str]],
     date_column: str,
     sample_size: int,
     random_state: int = 42,
@@ -297,6 +297,8 @@ def balance_undersample_time_series_trunc(
     time_unit_threshold: pd.Timedelta = DEFAULT_TIME_UNIT_THRESHOLD,
     **kwargs,
 ):
+    if id_columns is None:
+        id_columns = []
     # Convert date column to datetime
     dates_df = df[id_columns + [date_column]].copy()
     dates_df[date_column] = pd.to_datetime(dates_df[date_column], unit="ms")

{upgini-1.2.71a3832.dev12.dist-info → upgini-1.2.72.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: upgini
-Version: 1.2.71a3832.dev12
+Version: 1.2.72
 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-1.2.71a3832.dev12.dist-info → upgini-1.2.72.dist-info}/RECORD

@@ -1,12 +1,12 @@
-upgini/__about__.py,sha256=okx02f-XOrtKlpdtJnV-aHreGFvFkxZ5NQhd5zxvhMk,34
+upgini/__about__.py,sha256=-EK4ypqJTIRrg6g1P6PtLXT9vC4Vq7zblqFi389VgwA,23
 upgini/__init__.py,sha256=LXSfTNU0HnlOkE69VCxkgIKDhWP-JFo_eBQ71OxTr5Y,261
 upgini/ads.py,sha256=nvuRxRx5MHDMgPr9SiU-fsqRdFaBv8p4_v1oqiysKpc,2714
 upgini/dataset.py,sha256=aspri7ZAgwkNNUiIgQ1GRXvw8XQii3F4RfNXSrF4wrw,35365
 upgini/errors.py,sha256=2b_Wbo0OYhLUbrZqdLIx5jBnAsiD1Mcenh-VjR4HCTw,950
-upgini/features_enricher.py,sha256=DgWboHEhr5BQT87MaAo2iUtrhapP3iqczLeZtWLRkDs,206664
+upgini/features_enricher.py,sha256=Li1sPihWVkPUPcma8HRbPFwpCqd9V9d2p5zQUgkpdpU,206998
 upgini/http.py,sha256=RvzcShpDXssLs6ycGN8xilkKi8ZV9XGUrrk8bwdUzbw,43607
 upgini/metadata.py,sha256=Yd6iW2f7Wz6vUkg5uvR4xylN16ANnCKVKqAsAkap7p8,12354
-upgini/metrics.py,sha256=9AaQi7Yb22ZNnycUOAUpcP7TWF5Pfy_NGACcDj10aMs,38820
+upgini/metrics.py,sha256=a0bY4oTMb-MgB1yC1IuTcEtotKZxAxjgV_QV2Z4V8u4,38988
 upgini/search_task.py,sha256=EuCGp0iCWz2fpuJgN6M47aP_CtIi3Oq9zw78w0mkKiU,17595
 upgini/spinner.py,sha256=4iMd-eIe_BnkqFEMIliULTbj6rNI2HkN_VJ4qYe0cUc,1118
 upgini/version_validator.py,sha256=DvbaAvuYFoJqYt0fitpsk6Xcv-H1BYDJYHUMxaKSH_Y,1509
@@ -66,11 +66,11 @@ upgini/utils/postal_code_utils.py,sha256=5M0sUqH2DAr33kARWCTXR-ACyzWbjDq_-0mmEml
 upgini/utils/progress_bar.py,sha256=N-Sfdah2Hg8lXP_fV9EfUTXz_PyRt4lo9fAHoUDOoLc,1550
 upgini/utils/sklearn_ext.py,sha256=HpaNQaKJisgNE7IZ71n7uswxTj7kbPglU2G3s1sORAc,45042
 upgini/utils/sort.py,sha256=8uuHs2nfSMVnz8GgvbOmgMB1PgEIZP1uhmeRFxcwnYw,7039
-upgini/utils/target_utils.py,sha256=b1GzO8_gMcwXSZ2v98CY50MJJBzKbWHId_BJGybXfkM,16579
+upgini/utils/target_utils.py,sha256=P0cCVRaakWLydYwFjk3TEaQfr0p0hfsJCvKRD8qcxiE,16650
 upgini/utils/track_info.py,sha256=G5Lu1xxakg2_TQjKZk4b5SvrHsATTXNVV3NbvWtT8k8,5663
 upgini/utils/ts_utils.py,sha256=26vhC0pN7vLXK6R09EEkMK3Lwb9IVPH7LRdqFIQ3kPs,1383
 upgini/utils/warning_counter.py,sha256=-GRY8EUggEBKODPSuXAkHn9KnEQwAORC0mmz_tim-PM,254
-upgini-1.2.71a3832.dev12.dist-info/METADATA,sha256=8jmuNEDPwjc-Wa6Bds0FjYqYgqf3LFMYyRGUDy5DME8,49102
-upgini-1.2.71a3832.dev12.dist-info/WHEEL,sha256=1yFddiXMmvYK7QYTqtRNtX66WJ0Mz8PYEiEUoOUUxRY,87
-upgini-1.2.71a3832.dev12.dist-info/licenses/LICENSE,sha256=5RRzgvdJUu3BUDfv4bzVU6FqKgwHlIay63pPCSmSgzw,1514
-upgini-1.2.71a3832.dev12.dist-info/RECORD,,
+upgini-1.2.72.dist-info/METADATA,sha256=OpaT2gblO8qGzEJBNf36-dPwbedHPP93bX0fPAOMl38,49091
+upgini-1.2.72.dist-info/WHEEL,sha256=1yFddiXMmvYK7QYTqtRNtX66WJ0Mz8PYEiEUoOUUxRY,87
+upgini-1.2.72.dist-info/licenses/LICENSE,sha256=5RRzgvdJUu3BUDfv4bzVU6FqKgwHlIay63pPCSmSgzw,1514
+upgini-1.2.72.dist-info/RECORD,,