upgini 1.1.285__tar.gz → 1.1.285a3418.post1__tar.gz

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: upgini
-Version: 1.1.285
+Version: 1.1.285a3418.post1
 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-1.1.285a3418.post1/src/upgini/__about__.py

@@ -0,0 +1 @@
+__version__ = "1.1.285a3418-1"

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/autofe/all_operands.py

@@ -1,7 +1,7 @@
 from typing import Dict
 
 from upgini.autofe.binary import Add, Divide, Max, Min, Multiply, Sim, Subtract
-from upgini.autofe.date import DateDiff, DateDiffType2, DateListDiff, DateListDiffBounded
+from upgini.autofe.date import DateDiff, DateDiffType2, DateListDiff, DateListDiffBounded, DatePercentile
 from upgini.autofe.groupby import GroupByThenAgg, GroupByThenRank
 from upgini.autofe.operand import Operand
 from upgini.autofe.unary import Abs, Floor, Freq, Log, Residual, Sigmoid, Sqrt, Square
@@ -49,6 +49,7 @@ ALL_OPERANDS: Dict[str, Operand] = {
         DateListDiffBounded(diff_unit="Y", aggregation="count", lower_bound=30, upper_bound=45),
         DateListDiffBounded(diff_unit="Y", aggregation="count", lower_bound=45, upper_bound=60),
         DateListDiffBounded(diff_unit="Y", aggregation="count", lower_bound=60),
+        DatePercentile(),
     ]
 }
 

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/autofe/date.py

@@ -1,9 +1,9 @@
-from typing import Any, Optional, Union
+from typing import Any, Dict, List, Optional, Union
 
 import numpy as np
 import pandas as pd
 from pandas.core.arrays.timedeltas import TimedeltaArray
-from pydantic import BaseModel
+from pydantic import BaseModel, validator
 
 from upgini.autofe.operand import PandasOperand
 
@@ -27,6 +27,17 @@ class DateDiff(PandasOperand, DateDiffMixin):
     is_binary = True
     has_symmetry_importance = True
 
+    def get_params(self) -> Dict[str, Optional[str]]:
+        res = super().get_params()
+        res.update(
+            {
+                "diff_unit": self.diff_unit,
+                "left_unit": self.left_unit,
+                "right_unit": self.right_unit,
+            }
+        )
+        return res
+
     def calculate_binary(self, left: pd.Series, right: pd.Series) -> pd.Series:
         left = self._convert_to_date(left, self.left_unit)
         right = self._convert_to_date(right, self.right_unit)
@@ -42,6 +53,17 @@ class DateDiffType2(PandasOperand, DateDiffMixin):
     is_binary = True
     has_symmetry_importance = True
 
+    def get_params(self) -> Dict[str, Optional[str]]:
+        res = super().get_params()
+        res.update(
+            {
+                "diff_unit": self.diff_unit,
+                "left_unit": self.left_unit,
+                "right_unit": self.right_unit,
+            }
+        )
+        return res
+
     def calculate_binary(self, left: pd.Series, right: pd.Series) -> pd.Series:
         left = self._convert_to_date(left, self.left_unit)
         right = self._convert_to_date(right, self.right_unit)
@@ -64,6 +86,15 @@ class DateListDiff(PandasOperand, DateDiffMixin):
     has_symmetry_importance = True
     aggregation: str
 
+    def get_params(self) -> Dict[str, Optional[str]]:
+        res = super().get_params()
+        res.update(
+            {
+                "aggregation": self.aggregation,
+            }
+        )
+        return res
+
     def __init__(self, **data: Any) -> None:
         if "name" not in data:
             data["name"] = f"date_diff_{data.get('aggregation')}"
@@ -116,3 +147,55 @@ class DateListDiffBounded(DateListDiff):
     def _agg(self, x):
         x = x[(x >= (self.lower_bound or -np.inf)) & (x < (self.upper_bound or np.inf))]
         return super()._agg(x)
+
+
+class DatePercentile(PandasOperand):
+    name = "date_per"
+    is_binary = True
+    output_type = "float"
+
+    date_unit: Optional[str] = None
+    zero_month: Optional[int]
+    zero_year: Optional[int]
+    zero_bounds: Optional[List[float]]
+    step: int = 30
+
+    def get_params(self) -> Dict[str, Optional[str]]:
+        res = super().get_params()
+        res.update(
+            {
+                "date_unit": self.date_unit,
+                "zero_month": self.zero_month,
+                "zero_year": self.zero_year,
+                "zero_bounds": self.zero_bounds,
+                "step": self.step,
+            }
+        )
+        return res
+
+    @validator("zero_bounds", pre=True)
+    def validate_bounds(cls, value):
+        if value is None or isinstance(value, list):
+            return value
+        elif isinstance(value, str):
+            return value[1:-1].split(", ")
+
+    def calculate_binary(self, left: pd.Series, right: pd.Series) -> pd.Series:
+        # Assuming that left is a date column, right is a feature column
+        left = pd.to_datetime(left, unit=self.date_unit)
+        months = left.dt.month
+        years = left.dt.year
+
+        month_diffs = 12 * (years - (self.zero_year or 0)) + (months - (self.zero_month or 0))
+        bounds = month_diffs.apply(
+            lambda d: np.array(self.zero_bounds if self.zero_bounds is not None else []) + d * 30
+        )
+
+        return right.index.to_series().apply(lambda i: self.__perc(right[i], bounds[i]))
+
+    def __perc(self, f, bounds):
+        hit = np.where(f >= bounds)[0]
+        if hit.size > 0:
+            return np.max(hit) + 1
+        else:
+            return np.nan

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/autofe/feature.py

@@ -1,6 +1,6 @@
 import hashlib
 import itertools
-from typing import Dict, List, Optional, Tuple, Union
+from typing import Dict, List, Optional, Set, Tuple, Union
 
 import numpy as np
 import pandas as pd
@@ -16,6 +16,12 @@ class Column:
         self.data = data
         self.calculate_all = calculate_all
 
+    def get_display_name(self, cache: bool = True, shorten: bool = False, **kwargs) -> str:
+        return self.name
+
+    def set_op_params(self, params: Dict[str, str]) -> "Column":
+        return self
+
     def rename_columns(self, mapping: Dict[str, str]) -> "Column":
         self.name = self._unhash(mapping.get(self.name) or self.name)
         return self
@@ -69,19 +75,30 @@ class Feature:
         self.cached_display_name = cached_display_name
         self.alias = alias
 
-    def set_op_params(self, params: Dict[str, str]) -> "Feature":
+    def set_op_params(self, params: Optional[Dict[str, str]]) -> "Feature":
+        obj_dict = self.op.dict().copy()
+        obj_dict.update(params or {})
+        self.op = self.op.__class__.parse_obj(obj_dict)
         self.op.set_params(params)
+
+        for child in self.children:
+            child.set_op_params(params)
         return self
 
     def get_hash(self) -> str:
-        return hashlib.sha256("_".join([self.op.name] + [ch.name for ch in self.children]).encode("utf-8")).hexdigest()[
-            :8
-        ]
+        return hashlib.sha256(
+            "_".join([self.op.name] + [ch.get_display_name() for ch in self.children]).encode("utf-8")
+        ).hexdigest()[:8]
 
     def set_alias(self, alias: str) -> "Feature":
         self.alias = alias
         return self
 
+    def get_all_operand_names(self) -> Set[str]:
+        return {self.op.name}.union(
+            {n for f in self.children if isinstance(f, Feature) for n in f.get_all_operand_names()}
+        )
+
     def rename_columns(self, mapping: Dict[str, str]) -> "Feature":
         for child in self.children:
             child.rename_columns(mapping)

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/autofe/operand.py

@@ -25,8 +25,10 @@ class Operand(BaseModel):
         self.params = params
         return self
 
-    def get_params(self) -> Dict[str, str]:
-        return self.params
+    def get_params(self) -> Dict[str, Optional[str]]:
+        res = {"alias": self.alias}
+        res.update(self.params or {})
+        return res
 
 
 MAIN_COLUMN = "main_column"

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/features_enricher.py

@@ -423,7 +423,7 @@ class FeaturesEnricher(TransformerMixin):
 
             self.logger.info("Start fit")
 
-            self.__validate_search_keys(self.search_keys)
+            self.__validate_search_keys(self.search_keys, self.search_id)
 
             # Validate client estimator params
             self._get_client_cat_features(estimator, X, self.search_keys)
@@ -557,7 +557,7 @@ class FeaturesEnricher(TransformerMixin):
 
             self.logger.info("Start fit_transform")
 
-            self.__validate_search_keys(self.search_keys)
+            self.__validate_search_keys(self.search_keys, self.search_id)
 
             search_progress = SearchProgress(0.0, ProgressStage.START_FIT)
             if progress_callback is not None:
@@ -972,8 +972,6 @@ class FeaturesEnricher(TransformerMixin):
 
                     # 2 Fit and predict with KFold estimator on enriched tds
                     # and calculate final metric (and uplift)
-                    enriched_metric = None
-                    uplift = None
                     enriched_estimator = None
                     if set(fitting_X.columns) != set(fitting_enriched_X.columns):
                         self.logger.info(
@@ -994,15 +992,18 @@ class FeaturesEnricher(TransformerMixin):
                             has_date=has_date,
                         )
                         enriched_metric = enriched_estimator.cross_val_predict(fitting_enriched_X, enriched_y_sorted)
-                        if enriched_metric is None:
+                        if etalon_metric is None:
                             self.logger.warning(
                                 f"Enriched {metric} on train combined features is None (maybe all features was removed)"
                             )
                             enriched_estimator = None
+                            uplift = None
                         else:
                             self.logger.info(f"Enriched {metric} on train combined features: {enriched_metric}")
-                        if etalon_metric is not None and enriched_metric is not None:
                             uplift = (enriched_metric - etalon_metric) * multiplier
+                    else:
+                        enriched_metric = None
+                        uplift = None
 
                     train_metrics = {
                         self.bundle.get("quality_metrics_segment_header"): self.bundle.get(
@@ -1451,15 +1452,12 @@ class FeaturesEnricher(TransformerMixin):
             if len(decimal_columns_to_fix) > 0:
                 for col in decimal_columns_to_fix:
                     fitting_eval_X[col] = (
-                        fitting_eval_X[col]
-                        .astype("string").str
-                        .replace(",", ".", regex=False)
-                        .astype(np.float64)
+                        fitting_eval_X[col].astype("string").str.replace(",", ".", regex=False).astype(np.float64)
                     )
                     fitting_enriched_eval_X[col] = (
                         fitting_enriched_eval_X[col]
-                        .astype("string").str
-                        .replace(",", ".", regex=False)
+                        .astype("string")
+                        .str.replace(",", ".", regex=False)
                         .astype(np.float64)
                     )
 
@@ -2149,7 +2147,7 @@ class FeaturesEnricher(TransformerMixin):
             ]
             return excluded_features[feature_name_header].values.tolist()
 
-    def __validate_search_keys(self, search_keys: Dict[str, SearchKey], search_id: Optional[str] = None):
+    def __validate_search_keys(self, search_keys: Dict[str, SearchKey], search_id: Optional[str]):
         if (search_keys is None or len(search_keys) == 0) and self.country_code is None:
             if search_id:
                 self.logger.debug(f"search_id {search_id} provided without search_keys")
@@ -3278,10 +3276,16 @@ class FeaturesEnricher(TransformerMixin):
 
             descriptions = []
             for m in autofe_meta:
-                autofe_feature = Feature.from_formula(m.formula)
                 orig_to_hashed = {base_column.original_name: base_column.hashed_name for base_column in m.base_columns}
-                autofe_feature.rename_columns(orig_to_hashed)
-                autofe_feature.set_display_index(m.display_index)
+
+                autofe_feature = (
+                    Feature.from_formula(m.formula)
+                    .set_display_index(m.display_index)
+                    .set_alias(m.alias)
+                    .set_op_params(m.operator_params or {})
+                    .rename_columns(orig_to_hashed)
+                )
+
                 if autofe_feature.op.is_vector:
                     continue
 
@@ -3302,7 +3306,7 @@ class FeaturesEnricher(TransformerMixin):
                     description[f"Feature {feature_idx}"] = bc.hashed_name
                     feature_idx += 1
 
-                description["Function"] = autofe_feature.op.name
+                description["Function"] = ",".join(sorted(autofe_feature.get_all_operand_names()))
 
                 descriptions.append(description)
 

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/metadata.py

@@ -256,9 +256,11 @@ class BaseColumnMetadata(BaseModel):
 
 
 class GeneratedFeatureMetadata(BaseModel):
-    formula: str  # on hashed names
+    alias: Optional[str]
+    formula: str
     display_index: str
     base_columns: List[BaseColumnMetadata]
+    operator_params: Optional[Dict[str, str]]
 
 
 class ProviderTaskMetadataV2(BaseModel):

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/utils/custom_loss_utils.py

@@ -11,49 +11,46 @@ def get_runtime_params_custom_loss(
     runtime_parameters: RuntimeParameters,
     logger: Optional[logging.Logger] = None,
 ) -> RuntimeParameters:
-    if not loss:
-        return runtime_parameters
-
     if logger is None:
         logger = logging.getLogger()
-
-    selection_loss_reg = [
-        "regression",
-        "regression_l1",
-        "huber",
-        "poisson",
-        "quantile",
-        "mape",
-        "mean_absolute_percentage_error",
-        "gamma",
-        "tweedie",
-    ]
-    selection_loss_binary = ["binary"]
-    selection_loss_multi_clf = ["multiclass", "multiclassova", "multiclass_ova", "ova", "ovr"]
-    use_custom_loss = (
-        True
-        if (
-            (model_task_type == ModelTaskType.REGRESSION)
-            and (loss in selection_loss_reg)
-            or (model_task_type == ModelTaskType.BINARY)
-            and (loss in selection_loss_binary)
-            or (model_task_type == ModelTaskType.MULTICLASS)
-            and (loss in selection_loss_multi_clf)
+    if loss is not None:
+        selection_loss_reg = [
+            "regression",
+            "regression_l1",
+            "huber",
+            "poisson",
+            "quantile",
+            "mape",
+            "mean_absolute_percentage_error",
+            "gamma",
+            "tweedie",
+        ]
+        selection_loss_binary = ["binary"]
+        selection_loss_multi_clf = ["multiclass", "multiclassova", "multiclass_ova", "ova", "ovr"]
+        use_custom_loss = (
+            True
+            if (
+                (model_task_type == ModelTaskType.REGRESSION)
+                and (loss in selection_loss_reg)
+                or (model_task_type == ModelTaskType.BINARY)
+                and (loss in selection_loss_binary)
+                or (model_task_type == ModelTaskType.MULTICLASS)
+                and (loss in selection_loss_multi_clf)
+            )
+            else False
         )
-        else False
-    )
 
-    if use_custom_loss:
-        runtime_parameters.properties["lightgbm_params_preselection.objective"] = loss
-        runtime_parameters.properties["lightgbm_params_base.objective"] = loss
-        runtime_parameters.properties["lightgbm_params_segment.objective"] = loss
-        msg = bundle.get("loss_selection_info").format(loss)
-        logger.info(msg)
-        print(msg)
-    else:
-        msg = bundle.get("loss_selection_warn").format(loss, model_task_type)
-        logger.warning(msg)
-        print(msg)
+        if use_custom_loss:
+            runtime_parameters.properties["lightgbm_params_preselection.objective"] = loss
+            runtime_parameters.properties["lightgbm_params_base.objective"] = loss
+            runtime_parameters.properties["lightgbm_params_segment.objective"] = loss
+            msg = bundle.get("loss_selection_info").format(loss)
+            logger.info(msg)
+            print(msg)
+        else:
+            msg = bundle.get("loss_selection_warn").format(loss, model_task_type)
+            logger.warning(msg)
+            print(msg)
 
     return runtime_parameters
 

{upgini-1.1.285 → upgini-1.1.285a3418.post1}/src/upgini/version_validator.py

@@ -35,7 +35,7 @@ def validate_version(logger: logging.Logger):
         try:
             current_version = parse(__version__)
             latest_version = get_version("upgini")
-            if current_version < latest_version:
+            if current_version < latest_version:  # type: ignore
                 msg = bundle.get("version_warning").format(current_version, latest_version)
                 logger.warning(msg)
                 print(msg)

upgini-1.1.285/src/upgini/__about__.py

@@ -1 +0,0 @@
-__version__ = "1.1.285"