Walk through the whole pipeline, editing anything.
Upload one or more files, then move through cleaning, feature engineering,
and modeling one step at a time. Every step shows exactly what it's about to do and lets
you change it — pick a different action per column, add or skip engineered features,
drop columns, or run different models — before moving on.
Use this when you already have two files with the same columns —
one to train on, one to evaluate on — rather than two different datasets to join. The cleaning
plan, encoding, and feature engineering are fit on the train file only and applied to the test
file as-is; models are compared via cross-validation within the train file, then the winner gets
one final evaluation on the test file so you can see both scores side by side.
Have just one file? Upload it here and the next step lets you choose how it gets
split into a training portion and a held-out test portion — random, stratified (preserves class
proportions), or time-based (earliest rows train, latest rows test) — with a preview before
anything is committed.