lecrapaud 0.14.2__tar.gz → 0.14.3__tar.gz

{lecrapaud-0.14.2 → lecrapaud-0.14.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: lecrapaud
-Version: 0.14.2
+Version: 0.14.3
 Summary: Framework for machine and deep learning, with regression, classification and time series analysis
 License: Apache License
 Author: Pierre H. Gallet

{lecrapaud-0.14.2 → lecrapaud-0.14.3}/lecrapaud/api.py

@@ -119,7 +119,9 @@ class ExperimentEngine:
             os.makedirs(data_dir, exist_ok=True)
         else:
             if data is None:
-                raise ValueError("Either id or data must be provided")
+                raise ValueError(
+                    "Either id or data must be provided. Data can be a path to a folder containing trained models"
+                )
             self.experiment = create_experiment(data=data, **kwargs)
 
         # Set all kwargs as instance attributes

{lecrapaud-0.14.2 → lecrapaud-0.14.3}/lecrapaud/experiment.py

@@ -35,7 +35,7 @@ def create_experiment(
     groups = {}
     if group_column:
         groups["number_of_groups"] = data[group_column].nunique()
-        groups["list_of_groups"] = data[group_column].unique().tolist()
+        groups["list_of_groups"] = data[group_column].unique().tolist().sort()
 
     with get_db() as db:
         all_targets = Target.get_all(db=db)

{lecrapaud-0.14.2 → lecrapaud-0.14.3}/lecrapaud/feature_engineering.py

@@ -48,6 +48,7 @@ import pandas as pd
 import numpy as np
 from itertools import product
 import joblib
+import os
 
 from sklearn.compose import ColumnTransformer
 from sklearn.decomposition import PCA
@@ -382,9 +383,11 @@ class PreprocessFeature:
         return train, val, test
 
     def inference(self):
+        data = self.data
         # PCA
-        pcas = joblib.load(f"{self.preprocessing_dir}/pcas.pkl")
-        data, _ = self.add_pca_features(self.data, pcas=pcas)
+        if os.path.exists(f"{self.preprocessing_dir}/pcas.pkl"):
+            pcas = joblib.load(f"{self.preprocessing_dir}/pcas.pkl")
+            data, _ = self.add_pca_features(data, pcas=pcas)
 
         # Encoding
         transformer = joblib.load(f"{self.preprocessing_dir}/column_transformer.pkl")

{lecrapaud-0.14.2 → lecrapaud-0.14.3}/lecrapaud/feature_selection.py

@@ -313,6 +313,13 @@ class FeatureSelectionEngine:
             f"Final pre-selection: {len(features)} features below {corr_threshold}% out of {len(features_selected_list)} features, and rejected {len(features_correlated)} features, {100*len(features)/len(features_selected_list):.2f}% features selected"
         )
 
+        pd.Series(features).to_csv(
+            f"{self.feature_selection_dir}/features.csv",
+            index=True,
+            header=True,
+            index_label="ID",
+        )
+
         # analysis 2
         features_selected_by_every_methods_uncorrelated = list(
             set(features) & set(features_selected_by_every_methods)

{lecrapaud-0.14.2 → lecrapaud-0.14.3}/lecrapaud/model_selection.py

@@ -11,6 +11,7 @@ import joblib
 import glob
 from pathlib import Path
 import pickle
+from pydantic import BaseModel
 
 os.environ["COVERAGE_FILE"] = str(Path(".coverage").resolve())
 
@@ -1585,9 +1586,20 @@ def plot_confusion_matrix(y_true, y_pred):
     plt.show()
 
 
+class Threshold(BaseModel):
+    threshold: float
+    precision: float
+    recall: float
+    f1: float
+
+
+class Thresholds(BaseModel):
+    thresholds: dict[str, Threshold]
+
+
 def find_best_threshold(
     prediction: pd.DataFrame, metric: str = "recall", target_value: float | None = None
-) -> dict:
+) -> Thresholds:
     """
     General function to find best threshold optimizing recall, precision, or f1.
 
@@ -1599,7 +1611,7 @@ def find_best_threshold(
     - target_value (float | None): minimum acceptable value for the chosen metric.
 
     Returns:
-    - dict: {class_label: {'threshold', 'precision', 'recall', 'f1'}}
+    - Thresholds: {class_label: {'threshold', 'precision', 'recall', 'f1'}}
     """
     assert metric in {"recall", "precision", "f1"}, "Invalid metric"
     y_true = prediction["TARGET"]
@@ -1687,14 +1699,14 @@ def find_best_threshold(
 
 
 def apply_thresholds(
-    pred_proba: pd.DataFrame, threshold: dict | int | float, classes
+    pred_proba: pd.DataFrame, threshold: Thresholds | float, classes
 ) -> pd.DataFrame:
     """
     Apply thresholds to predicted probabilities.
 
     Parameters:
     - pred_proba (pd.DataFrame): Probabilities per class.
-    - threshold (float | dict): Global threshold (float) or per-class dict from `find_best_threshold`.
+    - threshold (Thresholds | float): Global threshold (float) or per-class dict from `find_best_threshold`.
     - classes (iterable): List or array of class labels (used for binary classification).
 
     Returns:
@@ -1817,8 +1829,6 @@ def plot_threshold(prediction, threshold, precision, recall):
 
 
 # OLD - to sort out
-
-
 def print_model_estimators(target_dir: str, model_name="linear"):
     """
     Look at a specific trained model

{lecrapaud-0.14.2 → lecrapaud-0.14.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "lecrapaud"
-version = "0.14.2"
+version = "0.14.3"
 description = "Framework for machine and deep learning, with regression, classification and time series analysis"
 authors = [
     {name = "Pierre H. Gallet"}