geocif 0.2.68__tar.gz → 0.2.69__tar.gz

{geocif-0.2.68/geocif.egg-info → geocif-0.2.69}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: geocif
-Version: 0.2.68
+Version: 0.2.69
 Summary: Models to visualize and forecast crop conditions and yields
 Home-page: https://ritviksahajpal.github.io/yield_forecasting/
 Author: Ritvik Sahajpal

{geocif-0.2.68 → geocif-0.2.69}/geocif/geocif.py

@@ -836,12 +836,13 @@ class Geocif:
 
             """ Feature selection and then Train """
             # Filter dataframe based on region and self.feature_names
+            breakpoint()
             df_region_train = self.df_train[mask_train]
             df_region_train = df_region_train[self.fixed_columns + common_columns]
             df_region_train.reset_index(drop=True, inplace=True)
             df_region_train = df_region_train.dropna(subset=[self.target_column])
 
-            self.X_train = df_region_train[self.feature_names]
+            self.X_train = df_region_train[self.feature_names + ["Region"]]
 
             # Drop any columns with NaNs except the lag yield columns
             lag_prefix = "t -"

{geocif-0.2.68 → geocif-0.2.69}/geocif/ml/feature_selection.py

@@ -65,7 +65,6 @@ def select_features(
     X_filtered : pd.DataFrame of selected features
     selected_features : list[str]
     """
-
     # copy original for multi-mode recursion
     X_clean = X.copy()
 
@@ -250,10 +249,30 @@ def select_features(
 
     elif method == "BorutaPy":
         from boruta import BorutaPy
-        sel = BorutaPy(forest, n_estimators="auto", random_state=42, verbose=2)
-        sel.fit(X_clean.values, y)
-        mask = sel.support_ | sel.support_weak_
-        selected = X_clean.columns[mask].tolist()
+        from collections import Counter
+        import itertools as it
+
+        region_selected = {}                         # {region: [features …]}
+        for region in tqdm(X_clean["Region"].unique(), desc="BorutaPy", leave=False):
+            idx = X_clean["Region"] == region
+            X_region = X_clean.loc[idx].drop(columns=["Region"])
+            y_region = y.loc[idx] if hasattr(y, "loc") else y[idx]
+
+            sel = BorutaPy(
+                estimator=forest,
+                n_estimators="auto",
+                random_state=42,
+                verbose=0          
+            )
+            sel.fit(X_region.values, y_region.values)
+
+            region_selected[region] = (
+                X_region.columns[sel.support_ | sel.support_weak_].tolist()
+            )
+
+        # ─── 3. keep features chosen in ≥ 2 regions ------------------------------
+        counts   = Counter(it.chain.from_iterable(region_selected.values()))
+        selected = [feat for feat, n in counts.items() if n >= 2]
 
     elif method == "Leshy":
         import arfs.feature_selection.allrelevant as arfsgroot

{geocif-0.2.68 → geocif-0.2.69/geocif.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: geocif
-Version: 0.2.68
+Version: 0.2.69
 Summary: Models to visualize and forecast crop conditions and yields
 Home-page: https://ritviksahajpal.github.io/yield_forecasting/
 Author: Ritvik Sahajpal

{geocif-0.2.68 → geocif-0.2.69}/setup.py

@@ -50,6 +50,6 @@ setup(
     test_suite="tests",
     tests_require=test_requirements,
     url="https://ritviksahajpal.github.io/yield_forecasting/",
-    version="0.2.68",
+    version="0.2.69",
     zip_safe=False,
 )