upgini 1.2.114a4__tar.gz → 1.2.114a6__tar.gz

{upgini-1.2.114a4 → upgini-1.2.114a6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: upgini
-Version: 1.2.114a4
+Version: 1.2.114a6
 Summary: Intelligent data search & enrichment for Machine Learning
 Project-URL: Bug Reports, https://github.com/upgini/upgini/issues
 Project-URL: Homepage, https://upgini.com/
@@ -38,7 +38,7 @@ Requires-Dist: python-bidi==0.4.2
 Requires-Dist: python-dateutil>=2.8.0
 Requires-Dist: python-json-logger>=3.3.0
 Requires-Dist: requests>=2.8.0
-Requires-Dist: scikit-learn>=1.3.0
+Requires-Dist: scikit-learn<1.8.0,>=1.3.0
 Requires-Dist: scipy>=1.10.0
 Requires-Dist: shap>=0.44.0
 Requires-Dist: xhtml2pdf<0.3.0,>=0.2.11
@@ -749,6 +749,36 @@ enricher.fit(
 )
 ```
 
+### Control feature stability with PSI parameters
+
+`FeaturesEnricher` supports Population Stability Index (PSI) calculation on eval_set to evaluate feature stability over time. You can control this behavior using stability parameters in `fit` and `fit_transform` methods:
+
+```python
+enricher = FeaturesEnricher(
+    search_keys={"registration_date": SearchKey.DATE}
+)
+
+# Control feature stability during fit
+enricher.fit(
+    X, y, 
+    stability_threshold=0.2,  # PSI threshold: features with PSI above this value will be dropped
+    stability_agg_func="max"  # Aggregation function for stability values: "max", "min", "mean"
+)
+
+# Same parameters work for fit_transform
+enriched_df = enricher.fit_transform(
+    X, y,
+    stability_threshold=0.1,   # Stricter threshold for more stable features
+    stability_agg_func="mean"  # Use mean aggregation instead of max
+)
+```
+
+**Stability parameters:**
+- `stability_threshold` (float, default=0.2): PSI threshold value. Features with PSI below this threshold will be excluded from the final feature set. Lower values mean stricter stability requirements.
+- `stability_agg_func` (str, default="max"): Function to aggregate PSI values across time intervals. Options: "max" (most conservative), "min" (least conservative), "mean" (balanced approach).
+
+**PSI (Population Stability Index)** measures how much feature distribution changes over time. Lower PSI values indicate more stable features, which are generally more reliable for production ML models.
+
 ### Use custom loss function in feature selection & metrics calculation
 
 `FeaturesEnricher` can be initialized with additional string parameter `loss`. 
@@ -769,20 +799,6 @@ enricher = FeaturesEnricher(
 enriched_dataframe.fit(X, y)
 ```
 
-### Return initial dataframe enriched with TOP external features by importance
-
-`transform` and `fit_transform` methods of `FeaturesEnricher` can be used with two additional parameters:
-- `importance_threshold`: float = 0 - only features with *importance >= threshold* will be added to the output dataframe
-- `max_features`: int  - only first TOP N features by importance will be returned, where *N = max_features*  
-
-And `keep_input=True` will keep all initial columns from search dataset X:  
-```python
-enricher = FeaturesEnricher(
-	search_keys={"subscription_activation_date": SearchKey.DATE}
-)
-enriched_dataframe.fit_transform(X, y, keep_input=True, max_features=2)
-```
-
 ### Exclude premium data sources from fit, transform and metrics calculation
 
 `fit`, `fit_transform`, `transform` and `calculate_metrics` methods of `FeaturesEnricher` can be used with parameter `exclude_features_sources` that allows to exclude Trial or Paid features from Premium data sources:

{upgini-1.2.114a4 → upgini-1.2.114a6}/README.md

@@ -703,6 +703,36 @@ enricher.fit(
 )
 ```
 
+### Control feature stability with PSI parameters
+
+`FeaturesEnricher` supports Population Stability Index (PSI) calculation on eval_set to evaluate feature stability over time. You can control this behavior using stability parameters in `fit` and `fit_transform` methods:
+
+```python
+enricher = FeaturesEnricher(
+    search_keys={"registration_date": SearchKey.DATE}
+)
+
+# Control feature stability during fit
+enricher.fit(
+    X, y, 
+    stability_threshold=0.2,  # PSI threshold: features with PSI above this value will be dropped
+    stability_agg_func="max"  # Aggregation function for stability values: "max", "min", "mean"
+)
+
+# Same parameters work for fit_transform
+enriched_df = enricher.fit_transform(
+    X, y,
+    stability_threshold=0.1,   # Stricter threshold for more stable features
+    stability_agg_func="mean"  # Use mean aggregation instead of max
+)
+```
+
+**Stability parameters:**
+- `stability_threshold` (float, default=0.2): PSI threshold value. Features with PSI below this threshold will be excluded from the final feature set. Lower values mean stricter stability requirements.
+- `stability_agg_func` (str, default="max"): Function to aggregate PSI values across time intervals. Options: "max" (most conservative), "min" (least conservative), "mean" (balanced approach).
+
+**PSI (Population Stability Index)** measures how much feature distribution changes over time. Lower PSI values indicate more stable features, which are generally more reliable for production ML models.
+
 ### Use custom loss function in feature selection & metrics calculation
 
 `FeaturesEnricher` can be initialized with additional string parameter `loss`. 
@@ -723,20 +753,6 @@ enricher = FeaturesEnricher(
 enriched_dataframe.fit(X, y)
 ```
 
-### Return initial dataframe enriched with TOP external features by importance
-
-`transform` and `fit_transform` methods of `FeaturesEnricher` can be used with two additional parameters:
-- `importance_threshold`: float = 0 - only features with *importance >= threshold* will be added to the output dataframe
-- `max_features`: int  - only first TOP N features by importance will be returned, where *N = max_features*  
-
-And `keep_input=True` will keep all initial columns from search dataset X:  
-```python
-enricher = FeaturesEnricher(
-	search_keys={"subscription_activation_date": SearchKey.DATE}
-)
-enriched_dataframe.fit_transform(X, y, keep_input=True, max_features=2)
-```
-
 ### Exclude premium data sources from fit, transform and metrics calculation
 
 `fit`, `fit_transform`, `transform` and `calculate_metrics` methods of `FeaturesEnricher` can be used with parameter `exclude_features_sources` that allows to exclude Trial or Paid features from Premium data sources:

{upgini-1.2.114a4 → upgini-1.2.114a6}/pyproject.toml

@@ -46,7 +46,7 @@ dependencies = [
     "python-dateutil>=2.8.0",
     "python-json-logger>=3.3.0",
     "requests>=2.8.0",
-    "scikit-learn>=1.3.0",
+    "scikit-learn>=1.3.0,<1.8.0",
     "scipy>=1.10.0",
     "python-bidi==0.4.2",
     "xhtml2pdf>=0.2.11,<0.3.0",

upgini-1.2.114a6/src/upgini/__about__.py

@@ -0,0 +1 @@
+__version__ = "1.2.114"

{upgini-1.2.114a4 → upgini-1.2.114a6}/src/upgini/dataset.py

@@ -25,7 +25,6 @@ from upgini.metadata import (
     AutoFEParameters,
     CVType,
     DataType,
-    FeaturesFilter,
     FileColumnMeaningType,
     FileColumnMetadata,
     FileMetadata,
@@ -37,8 +36,9 @@ from upgini.metadata import (
 )
 from upgini.resource_bundle import ResourceBundle, get_custom_bundle
 from upgini.search_task import SearchTask
+from upgini.utils.config import SampleConfig
 from upgini.utils.email_utils import EmailSearchKeyConverter
-from upgini.utils.sample_utils import SampleColumns, SampleConfig, sample
+from upgini.utils.sample_utils import SampleColumns, sample
 
 try:
     from upgini.utils.progress_bar import CustomProgressBar as ProgressBar
@@ -51,9 +51,6 @@ except Exception:
 class Dataset:
     MIN_ROWS_COUNT = 100
     MAX_ROWS = 200_000
-    IMBALANCE_THESHOLD = 0.6
-    MIN_TARGET_CLASS_ROWS = 100
-    MAX_MULTICLASS_CLASS_COUNT = 100
     MIN_SUPPORTED_DATE_TS = 946684800000  # 2000-01-01
     MAX_FEATURES_COUNT = 3500
     MAX_UPLOADING_FILE_SIZE = 268435456  # 256 Mb
@@ -73,6 +70,7 @@ class Dataset:
         cv_type: Optional[CVType] = None,
         date_column: Optional[str] = None,
         id_columns: Optional[List[str]] = None,
+        is_imbalanced: bool = False,
         random_state: Optional[int] = None,
         sample_config: Optional[SampleConfig] = None,
         rest_client: Optional[_RestClient] = None,
@@ -117,8 +115,9 @@ class Dataset:
         self.rest_client = rest_client
         self.random_state = random_state
         self.columns_renaming: Dict[str, str] = {}
-        self.imbalanced: bool = False
+        self.is_imbalanced: bool = False
         self.id_columns = id_columns
+        self.is_imbalanced = is_imbalanced
         self.date_column = date_column
         if logger is not None:
             self.logger = logger
@@ -239,8 +238,6 @@ class Dataset:
         else:
             train_segment = self.data
 
-        self.imbalanced = self.__is_imbalanced(train_segment)
-
         sample_columns = SampleColumns(
             ids=self.id_columns,
             date=self.date_column,
@@ -249,55 +246,19 @@ class Dataset:
         )
 
         self.data = sample(
-            train_segment if self.imbalanced else self.data,  # for imbalanced data we will be doing transform anyway
+            train_segment if self.is_imbalanced else self.data,  # for imbalanced data we will be doing transform anyway
             self.task_type,
             self.cv_type,
             self.sample_config,
             sample_columns,
             self.random_state,
-            balance=self.imbalanced,
+            balance=self.is_imbalanced,
             force_downsampling=force_downsampling,
             logger=self.logger,
             bundle=self.bundle,
             warning_callback=self.warning_callback,
         )
 
-    def __is_imbalanced(self, data: pd.DataFrame) -> bool:
-        if self.task_type is None or not self.task_type.is_classification():
-            return False
-
-        if self.task_type == ModelTaskType.BINARY and len(data) <= self.sample_config.binary_min_sample_threshold:
-            return False
-
-        count = len(data)
-        target_column = self.etalon_def_checked.get(FileColumnMeaningType.TARGET.value, TARGET)
-        target = data[target_column]
-        target_classes_count = target.nunique()
-
-        if target_classes_count > self.MAX_MULTICLASS_CLASS_COUNT:
-            msg = self.bundle.get("dataset_to_many_multiclass_targets").format(
-                target_classes_count, self.MAX_MULTICLASS_CLASS_COUNT
-            )
-            self.logger.warning(msg)
-            raise ValidationError(msg)
-
-        vc = target.value_counts()
-        min_class_value = vc.index[len(vc) - 1]
-        min_class_count = vc[min_class_value]
-
-        if min_class_count < self.MIN_TARGET_CLASS_ROWS:
-            msg = self.bundle.get("dataset_rarest_class_less_min").format(
-                min_class_value, min_class_count, self.MIN_TARGET_CLASS_ROWS
-            )
-            self.logger.warning(msg)
-            raise ValidationError(msg)
-
-        min_class_percent = self.IMBALANCE_THESHOLD / target_classes_count
-        min_class_threshold = min_class_percent * count
-
-        # If min class count less than 30% for binary or (60 / classes_count)% for multiclass
-        return bool(min_class_count < min_class_threshold)
-
     def __validate_dataset(self, validate_target: bool, silent_mode: bool):
         """Validate DataSet"""
         # self.logger.info("validating etalon")
@@ -537,9 +498,6 @@ class Dataset:
         return_scores: bool,
         extract_features: bool,
         accurate_model: Optional[bool] = None,
-        importance_threshold: Optional[float] = None,
-        max_features: Optional[int] = None,
-        filter_features: Optional[dict] = None,
         runtime_parameters: Optional[RuntimeParameters] = None,
         metrics_calculation: Optional[bool] = False,
         auto_fe_parameters: Optional[AutoFEParameters] = None,
@@ -548,28 +506,12 @@ class Dataset:
         search_customization = SearchCustomization(
             extractFeatures=extract_features,
             accurateModel=accurate_model,
-            importanceThreshold=importance_threshold,
-            maxFeatures=max_features,
             returnScores=return_scores,
             runtimeParameters=runtime_parameters,
             metricsCalculation=metrics_calculation,
         )
-        if filter_features:
-            if [
-                key
-                for key in filter_features
-                if key not in {"min_importance", "max_psi", "max_count", "selected_features"}
-            ]:
-                raise ValidationError(self.bundle.get("dataset_invalid_filter"))
-            feature_filter = FeaturesFilter(
-                minImportance=filter_features.get("min_importance"),
-                maxPSI=filter_features.get("max_psi"),
-                maxCount=filter_features.get("max_count"),
-                selectedFeatures=filter_features.get("selected_features"),
-            )
-            search_customization.featuresFilter = feature_filter
 
-        search_customization.runtimeParameters.properties["etalon_imbalanced"] = self.imbalanced
+        search_customization.runtimeParameters.properties["etalon_imbalanced"] = self.is_imbalanced
         if auto_fe_parameters is not None:
             search_customization.runtimeParameters.properties["feature_generation_params.ts.gap_days"] = (
                 auto_fe_parameters.ts_gap_days
@@ -624,9 +566,6 @@ class Dataset:
         extract_features: bool = False,
         accurate_model: bool = False,
         exclude_features_sources: Optional[List[str]] = None,
-        importance_threshold: Optional[float] = None,  # deprecated
-        max_features: Optional[int] = None,  # deprecated
-        filter_features: Optional[dict] = None,  # deprecated
         runtime_parameters: Optional[RuntimeParameters] = None,
         auto_fe_parameters: Optional[AutoFEParameters] = None,
         force_downsampling: bool = False,
@@ -643,9 +582,6 @@ class Dataset:
             return_scores=return_scores,
             extract_features=extract_features,
             accurate_model=accurate_model,
-            importance_threshold=importance_threshold,
-            max_features=max_features,
-            filter_features=filter_features,
             runtime_parameters=runtime_parameters,
             auto_fe_parameters=auto_fe_parameters,
         )