2 Commits

Author	SHA1	Message	Date
DevOps Engineer	82d6bdafba	HRT-43 — Test intégration ml_predictions_cache : zéro NULL hippodrome ... - Ajout tests/test_ml_cache_integrity.py : 7 tests integration vérifiant que hippodrome, race_label et heure ne sont pas NULL pour la date courante - Ajout marqueur 'integration' dans pytest.ini - Connexion DB en lecture seule (mode=ro) pour protection prod - Support variable d'env TEST_DATE et TURF_DB_PATH - Tests skippés proprement si job 19h30 n'a pas encore tourné - Validé sur les données 2026-04-26 : 7/7 PASSED (1005 lignes, 0 NULL) Co-Authored-By: Paperclip <noreply@paperclip.ing>	2026-04-27 14:26:46 +02:00
DevOps Engineer	6b762068fd	feat(ml): train ensemble model and generate benchmark report ... Results: - XGBoost (Optuna 100 trials): AUC=0.7856, Precision@3=0.5783 - LightGBM (Optuna 100 trials): AUC=0.7833, Precision@3=0.5736 - MLP (3 layers 256-128-64): AUC=0.7743, Precision@3=0.5643 - Ensemble (weighted voting): AUC=0.7840, Precision@3=0.5814 Baseline XGBoost: Precision@3=0.5287 Delta: +0.0527 (+5.3%) — DEPLOY threshold met (+5%) Latency: 35ms/race, 69ms/full-day (well under 200ms limit) SHAP: 31/43 features selected, top features: rang_cote, implied_prob, cote_direct, ratio_cote_field All 12 regression/latency tests passing. Co-Authored-By: Paperclip <noreply@paperclip.ing>	2026-04-25 19:10:41 +02:00