Merge pull request #186 from Leothosine/fix/issue-143-real-analytics-training-data

feat: replace synthetic training data with feature extraction from real analytics data

Author: BigBen-7

src/analytics_training_data.py

@@ -0,0 +1,59 @@
+"""Extract ML training features from real analytics data instead of synthetic data.
+
+Builds (X, y) arrays from TicketScan, TicketTransfer, and InvalidAttempt records
+stored in the database so the scalper-detection model trains on actual behaviour.
+"""
+from typing import Any, List, Tuple
+
+import numpy as np  # type: ignore[import-untyped]
+
+from src.analytics.models import AnalyticsStats, InvalidAttempt, TicketScan, TicketTransfer, get_session
+from sqlalchemy import func
+
+
+def extract_features_from_analytics() -> Tuple[Any, Any]:
+    """Return (X, y) arrays built from real analytics records.
+
+    Features per event_id:
+        0 - total scans
+        1 - invalid scan rate  (invalid / total scans, or 0)
+        2 - total transfers
+        3 - failed transfer rate (failed / total transfers, or 0)
+        4 - invalid attempt count
+
+    Label: 1 (suspicious) when invalid_scan_rate > 0.3 or failed_transfer_rate > 0.5,
+           0 otherwise.
+
+    Falls back to an empty array pair if no data is available.
+    """
+    session = get_session()
+    try:
+        # Aggregate per event_id from AnalyticsStats
+        stats = session.query(AnalyticsStats).all()
+
+        if not stats:
+            return np.empty((0, 5), dtype=float), np.empty((0,), dtype=int)
+
+        rows: List[List[float]] = []
+        labels: List[int] = []
+
+        for s in stats:
+            total_scans = max(s.scan_count, 1)
+            total_transfers = max(s.transfer_count, 1)
+            invalid_scan_rate = s.invalid_scan_count / total_scans
+            failed_transfer_rate = s.failed_transfer_count / total_transfers
+
+            features = [
+                float(s.scan_count),
+                invalid_scan_rate,
+                float(s.transfer_count),
+                failed_transfer_rate,
+                float(s.invalid_attempt_count),
+            ]
+            label = int(invalid_scan_rate > 0.3 or failed_transfer_rate > 0.5)
+            rows.append(features)
+            labels.append(label)
+
+        return np.array(rows, dtype=float), np.array(labels, dtype=int)
+    finally:
+        session.close()