"""The challenge's scoring, as in the Space's scoring.py. MAE_m = mean |predicted - actual| over the scored rows of metric m S_m = 1 - MAE_m / MAE_ref,m (ref = the 30-year-average baseline) Score = 100 * max(0, mean of the four S_m) Scored rows: valid years (10-day rule) of city-metrics that have a baseline (>= 10 valid years in the 30 years up to the origin). """ import numpy as np import pandas as pd from .common import METRICS KEY = ["city_id", "year", "metric"] def skill(mae: float, mae_ref: float) -> float: return 1.0 - mae / mae_ref def challenge_score(skills: dict[str, float]) -> float: """Score 0-100 from the four per-metric skills.""" assert set(skills) == set(METRICS) return 100.0 * max(0.0, float(np.mean([skills[m] for m in METRICS]))) def score_frames(submission: pd.DataFrame, answer_key: pd.DataFrame, reference: pd.DataFrame) -> dict: """Score a submission table the way the Space does (used to cross-check the array path). submission: city_id, year, metric, predicted_value answer_key: city_id, year, metric, actual_value, scored (bool) reference: the 30-year-average forecasts, same columns as submission """ truth = answer_key[answer_key["scored"].astype(bool)][KEY + ["actual_value"]] def mae(pred): rows = truth.merge(pred[KEY + ["predicted_value"]], on=KEY, how="left") if not np.isfinite(rows["predicted_value"]).all(): raise ValueError("scored rows without a finite prediction") return (rows["predicted_value"] - rows["actual_value"]).abs().groupby(rows["metric"]).mean() model, ref = mae(submission), mae(reference) skills = {m: skill(float(model[m]), float(ref[m])) for m in METRICS} return {"score": challenge_score(skills), "mae": {m: float(model[m]) for m in METRICS}, "mae_ref": {m: float(ref[m]) for m in METRICS}, "skill": skills, "n_rows": int(len(truth))}