feat: standardize feature engineering and push new production model

models/rossmann_production_model.pkl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:19f7950c78166df1b87f4b5e5db12b0782e6cd96a2f51502ddef7af41a225e7d
+oid sha256:59d3bda897c6d9678abfce7bd5b631183319efbdf275113d83463ad3d52f706e
 size 53873380

scripts/train_production_model.py

@@ -33,17 +33,16 @@ def run_production_training():
     df_feat = pipeline.run_feature_engineering(df_raw)
     
     # 3. Define Final Feature Set
-    # Include Store ID and Store Metadata
-    from sklearn.preprocessing import LabelEncoder
-    le = LabelEncoder()
-    for col in ['StoreType', 'Assortment']:
-        if col in df_feat.columns:
-            df_feat[col] = le.fit_transform(df_feat[col].astype(str))
+    # Consistent Encoding with app.py
+    if 'StoreType' in df_feat.columns:
+        df_feat['StoreType'] = df_feat['StoreType'].astype(str).map({'a':1, 'b':2, 'c':3, 'd':4}).fillna(0)
+    if 'Assortment' in df_feat.columns:
+        df_feat['Assortment'] = df_feat['Assortment'].astype(str).map({'a':1, 'b':2, 'c':3}).fillna(0)
 
     feature_cols = [
         'Store', 'DayOfWeek', 'Promo', 'StateHoliday', 'SchoolHoliday',
         'Year', 'Month', 'Day', 'IsWeekend', 'DayOfMonth',
-        'CompetitionDistance', 'CompetitionOpenTime', 'StoreType', 'Assortment'
+        'CompetitionDistance', 'StoreType', 'Assortment'
     ] + [c for c in df_feat.columns if 'fourier' in c or 'easter' in c]
     
     # 4. Final Training (using all available data to create the 'Gold' model)

src/features.py

@@ -22,7 +22,7 @@ class DateTransformation(FeatureEngineeringStrategy):
         df_transformed['Year'] = df_transformed[date_col].dt.year
         df_transformed['Month'] = df_transformed[date_col].dt.month
         df_transformed['Day'] = df_transformed[date_col].dt.day
-        df_transformed['DayOfWeek'] = df_transformed[date_col].dt.dayofweek
+        df_transformed['DayOfWeek'] = df_transformed[date_col].dt.dayofweek + 1
         df_transformed['IsWeekend'] = (df_transformed[date_col].dt.dayofweek >= 5).astype(int)
         df_transformed['DayOfMonth'] = df_transformed[date_col].dt.day
         return df_transformed