Update src/train_model.py

src/train_model.py

@@ -298,7 +298,7 @@ def generate_landslide_data(n: int = 4000):
     # ── Clean catalog spine ───────────────────────────────────────────────
     catalog = catalog.dropna(subset=["latitude", "longitude"])
     catalog["event_date"] = pd.to_datetime(
-        catalog["event_date"], format="%m-%d-%Y", errors="coerce"
+        catalog["event_date"], errors="coerce"
     )
     catalog = catalog.dropna(subset=["event_date"])
     print(f"[Landslide] After date clean: {len(catalog)} rows")
@@ -349,7 +349,7 @@ def generate_landslide_data(n: int = 4000):
         else:
             elev["aspect_index"] = 0.5
 
-    base = nearest_merge(base, elev, ["aspect_index"])
+    base = nearest_merge(base, elev, ["slope_degrees","aspect_index"])
 
     # ── Rainfall ──────────────────────────────────────────────────────────
     print("[Landslide] Merging rainfall...")
@@ -363,10 +363,11 @@ def generate_landslide_data(n: int = 4000):
     base = nearest_merge(base, rain_agg, ["rainfall_intensity_mmh"])
 
     # ── soil_type_index proxy ─────────────────────────────────────────────
+    slope_norm = np.clip(base["slope_degrees"] / 90.0, 0, 1)
     veg_norm  = np.clip(base["vegetation_cover_pct"] / 100.0, 0, 1)
     rain_norm = np.clip(base["rainfall_intensity_mmh"] / 200.0, 0, 1)
     base["soil_type_index"] = np.clip(                          # ← FIX 1: now saved
-        1.0 - (0.5 * (1 - veg_norm) + 0.5 * rain_norm), 0, 1
+        1.0 - (0.4 * slope_norm + 0.3 * (1 - veg_norm) + 0.3 * rain_norm), 0, 1
     )
 
     # ── Risk label ────────────────────────────────────────────────────────
@@ -395,9 +396,10 @@ def generate_landslide_data(n: int = 4000):
 
     base["risk_score"] = np.clip(
         0.35 * base["size_score"] +
-        0.30 * base["historical_landslide_freq"] +
-        0.25 * fatality_norm +
-        0.10 * (1 - veg_norm) +
+        0.25 * base["historical_landslide_freq"] +
+        0.20 * fatality_norm +
+        0.15 * slope_norm +
+        0.05 * (1 - veg_norm) +
         np.random.normal(0, 0.02, len(base)),
         0.0, 1.0
     )
@@ -408,7 +410,7 @@ def generate_landslide_data(n: int = 4000):
 
     # ── Final feature matrix ──────────────────────────────────────────────
     features = [
-        "rainfall_intensity_mmh", "soil_type_index",
+        "slope_degrees","rainfall_intensity_mmh", "soil_type_index",
         "vegetation_cover_pct", "seismic_activity_index",
         "distance_to_fault_km", "aspect_index", "historical_landslide_freq",
     ]