Fix/change prediction categories

blocksnet/analysis/land_use/prediction/schemas.py

@@ -1,12 +1,13 @@
 import pandas as pd
+from pandera import Field
 from pandera.typing import Series
 from shapely import MultiPolygon, Polygon
 from blocksnet.enums import LandUseCategory
 from blocksnet.utils.validation import GdfSchema, LandUseSchema
 
 
 class BlocksInputSchema(GdfSchema):
-    category: Series
+    category: Series = Field(nullable=True)
 
     @classmethod
     def _geometry_types(cls):
@@ -20,7 +21,22 @@ def _before_validate(cls, df: pd.DataFrame) -> pd.DataFrame:
             def parse_category(c):
                 if isinstance(c, LandUseCategory):
                     return c
-                return LandUseCategory(c.lower())
+                if isinstance(c, str):
+                    s = c.strip()
+                    # try value (exact), then case-insensitive by value, then by name
+                    try:
+                        return LandUseCategory(s)
+                    except Exception:
+                        pass
+                    try:
+                        return next(v for v in LandUseCategory if v.value.lower() == s.lower())
+                    except Exception:
+                        pass
+                    try:
+                        return LandUseCategory[s.upper()]
+                    except Exception:
+                        pass
+                return None
 
             df["category"] = df["category"].map(parse_category)
 

blocksnet/enums/land_use_category.py

@@ -1,15 +1,18 @@
 from enum import Enum
 from blocksnet.enums import LandUse
-
 class LandUseCategory(Enum):
-    """Enumeration class representing different categories of land use.
-
-    This enum defines three main categories of land use: urban, non-urban, and industrial.
-    It provides methods to convert between LandUse objects and LandUseCategory instances.
+    """High-level land use categories used for prediction.
+    Supported categories:
+    - INDUSTRIAL
+    - RECREATION
+    - BUSINESS
+    - RESIDENTIAL
     """
-    URBAN = "urban"
-    NON_URBAN = "non_urban"
-    INDUSTRIAL = "industrial"
+    RESIDENTIAL = "RESIDENTIAL"
+    BUSINESS = "BUSINESS"
+    RECREATION = "RECREATION"
+    INDUSTRIAL = "INDUSTRIAL"
+
     _REVERSE_MAP = None
 
     @classmethod
@@ -40,11 +43,12 @@ def to_land_use(self) -> set[LandUse]:
         return LandUseCategory._REVERSE_MAP.get(self, set())
 
 LU_MAPPING = {
-    LandUse.RESIDENTIAL: LandUseCategory.URBAN,
-    LandUse.BUSINESS: LandUseCategory.URBAN,
-    LandUse.RECREATION: LandUseCategory.NON_URBAN,
-    LandUse.TRANSPORT: LandUseCategory.NON_URBAN,
-    LandUse.SPECIAL: LandUseCategory.NON_URBAN,
-    LandUse.AGRICULTURE: LandUseCategory.NON_URBAN,
+    LandUse.RESIDENTIAL: LandUseCategory.RESIDENTIAL,
+    LandUse.BUSINESS: LandUseCategory.BUSINESS,
+    LandUse.RECREATION: LandUseCategory.RECREATION,
+    LandUse.AGRICULTURE: LandUseCategory.RECREATION,
+    LandUse.SPECIAL: LandUseCategory.RECREATION,
     LandUse.INDUSTRIAL: LandUseCategory.INDUSTRIAL,
-}
+    LandUse.TRANSPORT: LandUseCategory.INDUSTRIAL,
+    # Other LandUse values (e.g., TRANSPORT, SPECIAL, AGRICULTURE) map to None
+}

examples/analysis/land_use/artifacts/meta.json

@@ -0,0 +1 @@
+{"strategy_cls": "SKLearnVotingClassificationStrategy", "model_cls": "VotingClassifier", "model_params": {"voting": "soft", "n_jobs": -1}}