Add CI and shadow calibration reporting

2026-03-20 20:59:30 +08:00
parent 25ab512371
commit 1c84893bed
39 changed files with 149223 additions and 3277 deletions
@@ -0,0 +1,152 @@
+import argparse
+import json
+import os
+import sys
+
+PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+if PROJECT_ROOT not in sys.path:
+    sys.path.insert(0, PROJECT_ROOT)
+
+from src.analysis.deb_algorithm import load_history, save_history  # noqa: E402
+from src.analysis.probability_calibration import (  # noqa: E402
+    ENGINE_MODE_EMOS_SHADOW,
+    apply_probability_calibration,
+    build_probability_features,
+)
+
+
+def _sample_to_features(sample):
+    peak_flag = sample.get("peak_flag")
+    if peak_flag == 1.0:
+        peak_status = "past"
+    elif peak_flag == 0.5:
+        peak_status = "in_window"
+    else:
+        peak_status = "before"
+    return build_probability_features(
+        city_name=sample.get("city") or "",
+        raw_mu=sample.get("raw_mu"),
+        raw_sigma=sample.get("raw_sigma"),
+        deb_prediction=sample.get("deb_prediction"),
+        ens_data={
+            "median": sample.get("ens_median"),
+            "p10": None,
+            "p90": None,
+        },
+        current_forecasts={},
+        max_so_far=None,
+        peak_status=peak_status,
+        local_hour_frac=None,
+    )
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Backfill shadow probability snapshots into daily records.")
+    parser.add_argument(
+        "--history-file",
+        default=os.path.join(PROJECT_ROOT, "data", "daily_records.json"),
+    )
+    parser.add_argument(
+        "--training-samples",
+        default=os.path.join(
+            PROJECT_ROOT,
+            "artifacts",
+            "probability_calibration",
+            "training_samples.json",
+        ),
+    )
+    parser.add_argument(
+        "--calibration-file",
+        default=os.path.join(
+            PROJECT_ROOT,
+            "artifacts",
+            "probability_calibration",
+            "default.json",
+        ),
+    )
+    args = parser.parse_args()
+
+    history = load_history(args.history_file)
+    with open(args.training_samples, "r", encoding="utf-8") as fh:
+        training_payload = json.load(fh)
+
+    updated = 0
+    touched = 0
+
+    for sample in training_payload.get("samples") or []:
+        city = str(sample.get("city") or "").strip().lower()
+        date_str = str(sample.get("date") or "").strip()
+        if not city or not date_str:
+            continue
+        record = ((history.get(city) or {}).get(date_str) or {})
+        if not isinstance(record, dict) or not record:
+            continue
+
+        legacy_distribution = [
+            {"value": row.get("v"), "probability": row.get("p")}
+            for row in (record.get("prob_snapshot") or [])
+            if isinstance(row, dict) and row.get("v") is not None
+        ]
+        if not legacy_distribution:
+            continue
+
+        calibration = apply_probability_calibration(
+            city_name=city,
+            temp_symbol="°F" if city in {"atlanta", "chicago", "dallas", "miami", "new york", "seattle"} else "°C",
+            raw_mu=sample.get("raw_mu"),
+            raw_sigma=sample.get("raw_sigma"),
+            max_so_far=None,
+            legacy_distribution=legacy_distribution,
+            features=_sample_to_features(sample),
+            calibration_path=args.calibration_file,
+            mode=ENGINE_MODE_EMOS_SHADOW,
+        )
+
+        shadow_distribution = calibration.get("shadow_distribution") or []
+        compact_shadow = [
+            {
+                "v": int(row.get("value")),
+                "p": round(float(row.get("probability") or 0.0), 3),
+            }
+            for row in shadow_distribution[:4]
+            if row.get("value") is not None
+        ]
+        compact_calibration = {
+            "mode": calibration.get("mode"),
+            "engine": calibration.get("engine"),
+            "version": calibration.get("calibration_version"),
+            "source": calibration.get("calibration_source"),
+            "raw_mu": calibration.get("raw_mu"),
+            "raw_sigma": calibration.get("raw_sigma"),
+            "calibrated_mu": calibration.get("calibrated_mu"),
+            "calibrated_sigma": calibration.get("calibrated_sigma"),
+        }
+
+        touched += 1
+        if (
+            record.get("shadow_prob_snapshot") == compact_shadow
+            and record.get("probability_calibration") == compact_calibration
+        ):
+            continue
+
+        record["shadow_prob_snapshot"] = compact_shadow
+        record["probability_calibration"] = compact_calibration
+        history[city][date_str] = record
+        updated += 1
+
+    save_history(args.history_file, history)
+    print(
+        json.dumps(
+            {
+                "samples_seen": len(training_payload.get("samples") or []),
+                "records_considered": touched,
+                "records_updated": updated,
+            },
+            ensure_ascii=False,
+            indent=2,
+        )
+    )
+
+
+if __name__ == "__main__":
+    main()