# GFIS Dataset Catalogue

Status: scaffold created; source verification and licensing review are required before final experiments.

| Dataset ID | Source class | Local status | Intended role | Required next action |
|---|---|---|---|---|
| `gfis-synthetic-raw` | GFIS synthetic/literature-inspired | Copied | Baseline training and controlled stress tests | Record generator version, equations, units, seed, checksum, and limitations |
| `gfis-synthetic-processed` | Derived GFIS dataset | Copied | Current baseline model input | Rebuild from raw using frozen pipeline and verify leakage-free transformations |
| `gfis-soft-sensor-extract` | Derived GFIS evidence | Copied | VFA/ALK and stability demonstration | Trace every field to source/generator and document threshold provenance |
| `kaggle-ad-*` | Kaggle public dataset candidate | Not yet acquired | External validation or transfer study | Select authoritative dataset, verify owner/license/citation, record download URL/version, and build adapter |
| `mendeley-ad-*` | Mendeley Data candidate | Not yet acquired | Public research-data validation | Select dataset, capture DOI/version/license/citation, and build adapter |
| `adm1-benchmark-*` | ADM1 benchmark/simulation reference | Not yet acquired | Mechanistic benchmark and transition experiments | Select citable implementation/data, record equations/units/license, and distinguish benchmark simulation from plant data |

Each final catalogue entry must include: title, authors/owner, landing-page URL or DOI, version/date, license, citation, acquisition date, redistribution permission, file hashes, row count, time coverage, sampling frequency, variables and units, target definition, missingness, preprocessing, split boundaries, known limitations, and experiments using it.