v3 · Colab run · part 7 of 7

Submission

The notebook exactly as it ran on Colab's T4 GPU, with its real outputs.

Left: what & whyRight: the code and its real output

7. Submission

submission.csv = the best scenario. The decoder is re-tuned on all out-of-fold predictions, then applied to the test predictions. The 2nd and 3rd best are saved too (submission_rank2.csv, submission_rank3.csv). Pick by CV, not by the public leaderboard: the public part is only a slice of the test set.

prog.start("write submissions")
ranked = sorted(rows, key=lambda r: -r["score"][0])
for i, r in enumerate(ranked[:3]):
    o, t = r["out"]
    prm = r["fit"](o, y, W_tr, MT_tr)
    pred = r["apply"](t, prm, MT_te)
    fname = "submission.csv" if i == 0 else f"submission_rank{i + 1}.csv"
    pd.DataFrame({"ID": test["space_id"], "price_tier": pred.astype(int) + 1}).to_csv(WORK / fname, index=False)
    print(f"{fname:24s} ← {r['scenario']} | {r['decoder']} | CV (test-weighted) {r['score'][0]:.4f}")
prog.finish()
pd.read_csv(WORK / "submission.csv").price_tier.value_counts().sort_index()

▶ write submissions   [████████████████████████████░░ 20/21 (95%)]
submission.csv           ← stack | A class weights | CV (test-weighted) 0.7214
submission_rank2.csv     ← opt_blend + xgb_reg (a=0.7) | B cut-points | CV (test-weighted) 0.7199
submission_rank3.csv     ← opt_blend + xgb_reg (a=0.5) | B cut-points | CV (test-weighted) 0.7198
✓ write submissions: took 0.2 min | [██████████████████████████████ 21/21 (100%)] | elapsed 36.0 min | ETA ~-

	count
price_tier
1	25533
2	37412
3	26906
4	24687

dtype: int64