Modelli addestrati, metriche e segnali previsionali
| ID | #7 | Strategia | ZAP |
|---|---|---|---|
| Nome | ML V2 random_forest_ranked | Tipo | random_forest |
| Versione | v20260704_082124 | Status | completed |
| Target | target_outperform | Orizzonte | 60 giorni |
| Train | 27/05/2016 → 19/04/2023 | Validation | 20/04/2023 → 10/10/2024 |
| Test | 11/10/2024 → 07/04/2026 | Artifact | /home/zymail/web/sei.zymail.net/python/models/ml_v2_random_forest_ranked_strategy_1_20260704_082123.joblib |
| Periodo | Accuracy | Baseline | AUC | Precision | Recall | F1 | Top20 accuracy | Top20 alpha | Bottom20 alpha |
|---|---|---|---|---|---|---|---|---|---|
| Train | 55,96% | 51,24% | 0,5885 | 58,65% | 47,68% | 52,60% | 61,52% | 5,05% | -1,41% |
| Validation | 56,20% | 47,77% | 0,5876 | 54,79% | 47,58% | 50,93% | 60,94% | 4,45% | -2,78% |
| Test | 53,35% | 45,68% | 0,5360 | 48,87% | 46,16% | 47,48% | 49,10% | 1,57% | -1,82% |
Il modello è utile solo se supera la baseline del test. Baseline = percentuale naturale di casi che battono il benchmark.
AUC misura la capacità di ordinare bene i titoli. Sopra 0,50 c’è segnale; sopra 0,55 è già interessante per MVP.
È la metrica più operativa: indica quanto rende il gruppo migliore scelto dal modello rispetto al benchmark.
Il modello è ancora sperimentale. Serve aumentare l’universo titoli e fare walk-forward prima di usarlo con capitale reale.
| Feature | Importanza | Barra |
|---|---|---|
| atr_14_rank | 0,453568 | |
| volatility_120d_rank | 0,125346 | |
| old_risk_score_rank | 0,048961 | |
| volatility_60d_rank | 0,045831 | |
| volatility_inverse_rank | 0,045135 | |
| momentum_12m_rank | 0,043799 | |
| return_252d_rank | 0,040240 | |
| volatility_20d_rank | 0,028018 | |
| macd_signal_rank | 0,020682 | |
| price_vs_ma_200_rank | 0,018125 | |
| return_120d_rank | 0,016956 | |
| macd_rank | 0,016810 | |
| momentum_3m_rank | 0,015520 | |
| momentum_6m_rank | 0,015037 | |
| not_too_extended | 0,014309 | |
| old_momentum_score_rank | 0,013317 | |
| return_60d_rank | 0,012683 | |
| price_vs_ma_50_rank | 0,008726 | |
| old_composite_score_rank | 0,005035 | |
| momentum_1m_rank | 0,004312 | |
| return_20d_rank | 0,003582 | |
| rsi_distance_50_rank | 0,002007 | |
| rsi_14_rank | 0,001056 | |
| return_5d_rank | 0,000696 | |
| volume_ratio_20d_rank | 0,000168 |
| ID | Modello | Tipo | Accuracy | AUC | F1 | Status | Creato |
|---|---|---|---|---|---|---|---|
| #14 | ML V2 logistic_regression | logistic_regression | 52,74% | 0,5373 | 50,62% | completed | 2026-09-26 08:47:18 |
| #13 | ML V2 logistic_regression | logistic_regression | 52,69% | 0,5367 | 50,78% | completed | 2026-09-19 08:52:04 |
| #12 | ML V2 logistic_regression | logistic_regression | 52,54% | 0,5344 | 50,80% | completed | 2026-09-12 08:50:14 |
| #11 | ML V2 logistic_regression | logistic_regression | 52,51% | 0,5329 | 50,83% | completed | 2026-09-05 08:43:10 |
| #10 | ML V2 random_forest_ranked | random_forest | 53,20% | 0,5353 | 48,38% | completed | 2026-07-25 08:43:17 |
| #9 | ML V2 random_forest_ranked | random_forest | 53,27% | 0,5349 | 47,87% | completed | 2026-07-18 08:24:56 |
| #8 | ML V2 random_forest_ranked | random_forest | 53,21% | 0,5349 | 47,52% | completed | 2026-07-11 08:22:19 |
| #7 | ML V2 random_forest_ranked | random_forest | 53,35% | 0,5360 | 47,48% | completed | 2026-07-04 08:21:24 |
| #6 | ML V2 logistic_regression | logistic_regression | 53,24% | 0,5398 | 51,39% | completed | 2026-06-27 08:25:48 |
| #5 | ML V2 logistic_regression | logistic_regression | 53,43% | 0,5423 | 51,65% | completed | 2026-06-20 10:21:25 |
| #4 | ML V2 logistic_regression | logistic_regression | 53,44% | 0,5413 | 51,72% | completed | 2026-06-13 10:21:10 |
| #3 | ML V2 logistic_regression | logistic_regression | 53,61% | 0,5436 | 51,98% | completed | 2026-06-06 12:21:25 |
| #2 | ML V2 logistic_regression | logistic_regression | 53,80% | 0,5448 | 52,24% | completed | 2026-05-30 12:20:56 |
| #1 | ML V2 logistic_regression | logistic_regression | 53,85% | 0,5449 | 52,35% | completed | 2026-05-27 20:05:16 |