| """Time-based splits that mirror the official evaluation. |
| |
| Evaluation = each user's FIRST event at/after the cutoff. So for a cutoff C: |
| fit events = all train events with ts < C (labels usable for aggregates) |
| target/val = for each user, the first train event with C <= ts < C+W (W=None -> until end of train) |
| Whole events always stay together (split is by event, rows follow their event). |
| """ |
| import polars as pl |
|
|
| J20 = 1737331200 |
| DAY = 86400 |
| C_MAIN = J20 + 5 * DAY |
| C_SECOND = J20 + 6 * DAY |
| C_FINAL = J20 + 7 * DAY |
|
|
| def fit_events(TE: pl.DataFrame, C: int) -> pl.DataFrame: |
| return TE.filter(pl.col('ts') < C) |
|
|
| def first_events(TE: pl.DataFrame, C: int, W: int | None = None) -> pl.DataFrame: |
| """First event per user in [C, C+W). Ties on ts broken by eid (deterministic).""" |
| m = pl.col('ts') >= C |
| if W is not None: |
| m = m & (pl.col('ts') < C + W) |
| return TE.filter(m).sort('ts', 'eid').group_by('u', maintain_order=True).first() |
|
|
| def split(TE: pl.DataFrame, TR: pl.DataFrame, C: int, W: int | None = None): |
| """Return (fit_ev, fit_rows, val_ev, val_rows).""" |
| fe = fit_events(TE, C) |
| ve = first_events(TE, C, W) |
| fr = TR.join(fe.select('eid'), on='eid', how='semi') |
| vr = TR.join(ve.select('eid'), on='eid', how='semi') |
| return fe, fr, ve, vr |
|
|
| def train_fold_cutoffs(C_val: int): |
| """Daily cutoffs strictly before C_val used to build model-training rows (each fold: first event/user in [Ck, Ck+1d)).""" |
| return [J20 + k * DAY for k in range(1, (C_val - J20) // DAY)] |
|
|
