protostar-climate-2026 / code /forecast_v2.py
Tonic's picture
Protostar Climate submission: v2 forecast + v3/DCPP amendment audit (29 gated candidates)
49f3de8 verified
Raw History Blame Contribute Delete
2.54 kB
"""Generate the v2 submission.csv (adopted after Phase-5 confirmation).
Package (selected on tune-era origins 1990-2000, confirmed on 2001-2011):
- annual_avg_temp: median of {L15+national-40yr-trend, L15+regional-40yr-trend,
L15+national-40yr-trend on step-adjusted series}
- cdd: 50/50 blend of CDD delta method (30 historical years, median) and the
linear L15+national-40yr-trend CDD model
- total_precip_in, hottest_day_temp: 30-year mean (all contested candidates
were rejected in Phase 4)
Fitted on data <= 2018 only; predicts 2019-2025.
"""
from __future__ import annotations
import sys
from pathlib import Path
import numpy as np
import pandas as pd
sys.path.insert(0, str(Path(__file__).parent))
from dataset import METRICS
from leakage import assert_no_forbidden_artifacts
from run_v2_confirm import build_packages
ORIGIN = 2018
TARGET_YEARS = list(range(2019, 2026))
def main() -> None:
root = Path(__file__).resolve().parent.parent
assert_no_forbidden_artifacts(Path(__file__).parent)
panel, packages = build_packages()
assert int(panel.years.max()) <= ORIGIN
frames = []
for metric, cands in packages.items():
fc = cands["v2"](panel, ORIGIN) # (210, 7)
if not np.isfinite(fc).all():
raise RuntimeError(f"{metric}: non-finite forecasts")
frames.append(pd.DataFrame({
"city_id": np.repeat(panel.cities, 7),
"year": np.tile(TARGET_YEARS, len(panel.cities)),
"metric": metric,
"predicted_value": fc.ravel(),
}))
out = pd.concat(frames, ignore_index=True)
expected = {(c, y, m) for c in panel.cities for y in TARGET_YEARS for m in METRICS}
got = set(zip(out["city_id"], out["year"], out["metric"]))
if got != expected:
raise RuntimeError(f"grid mismatch: missing {len(expected - got)}, extra {len(got - expected)}")
out = out.sort_values(["city_id", "year", "metric"]).reset_index(drop=True)
v1_path = root / "outputs" / "submission_v1.csv"
sub_path = root / "submission.csv"
if sub_path.exists() and not v1_path.exists():
v1_path.write_bytes(sub_path.read_bytes())
print(f"backed up v1 to {v1_path}")
out.to_csv(sub_path, index=False)
out.to_csv(root / "outputs" / "submission_v2.csv", index=False)
print(f"wrote {sub_path}: {len(out)} rows")
print(out.groupby("metric")["predicted_value"].describe().round(2).to_string())
if __name__ == "__main__":
main()