Implement handle() function

Author: AAdewunmi

policylens/apps/claims/management/commands/train_completeness_model.py

@@ -0,0 +1,39 @@
+# path: policylens/apps/claims/management/commands/train_completeness_model.py
+"""
+Train the completeness classifier and save a versioned model bundle.
+
+This command:
+- Reads a synthetic dataset CSV (or any CSV matching the contract)
+- Trains a lightweight model
+- Saves artefacts to artifacts/ml/<version>/
+"""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+from django.core.management.base import BaseCommand
+
+from apps.claims.ml.train import train_from_csv
+
+
+class Command(BaseCommand):
+    """Train the completeness model."""
+
+    help = "Train completeness classifier from a contract-aligned CSV dataset."
+
+    def add_arguments(self, parser) -> None:
+        parser.add_argument("--csv", required=True, help="Input dataset path.")
+        parser.add_argument("--version", required=True, help="Model version folder name, e.g. v1_2026_01_13.")
+        parser.add_argument("--threshold", type=float, default=0.6, help="Threshold for likely incomplete.")
+        parser.add_argument("--seed", type=int, default=42, help="Deterministic training seed.")
+
+    def handle(self, *args, **options) -> None:
+        csv_path = Path(options["csv"])
+        version = str(options["version"])
+        threshold = float(options["threshold"])
+        seed = int(options["seed"])
+
+        result = train_from_csv(csv_path=csv_path, model_version=version, threshold=threshold, random_seed=seed)
+        self.stdout.write(self.style.SUCCESS(f"Trained model {result.model_version}"))
+        self.stdout.write(self.style.SUCCESS(f"Metrics: {result.metrics}"))