Catalog policy
Methodology and Data Quality
IoTDataset.com is a discovery catalog. It records where a dataset comes from and displays available licence, citation, preview, and generation metadata. A catalog record does not certify the underlying dataset.
Last updated: July 29, 2026
1. Scope of the catalog
External-source records point to publishers such as data repositories, universities, public agencies, and project websites. IoTDataset.com normally stores descriptive metadata and a source link rather than a copy of the publisher's files.
IoTSyn-generated records describe synthetic data produced by a named IoTSyn engine version. Where available, the record also exposes the run identifier, citation, licence, row and column counts, sample rows, and companion artifacts.
2. Meaning of record fields
| Field | Interpretation |
|---|---|
| Source link | The page or file recorded as the source. Its availability and contents may change. |
| Licence | The licence URL recorded from the publisher or generator. If absent, no reuse right is inferred. |
| Citation | A publisher-supplied or generator-supplied citation when recorded. Otherwise the page offers only a citation for the catalog record. |
| Description | A catalog summary for discovery. Publisher documentation controls when the two differ. |
| Preview | A limited sample or stored table, not proof of the complete file's schema or integrity. |
| Engine version | The IoTSyn version recorded for a generated run; it must be considered with configuration, seed, dependencies, and environment. |
3. Review workflow
- Preserve the public record identifier and slug so existing references continue to resolve.
- Confirm that the source link identifies the intended dataset or publication.
- Record a licence only when the publisher or generator states one; otherwise mark it as not recorded.
- Prefer a publisher-provided citation or structured BibTeX record. Do not invent missing authors, titles, or DOIs.
- For generated data, record the engine version and retain available run artifacts and validation outputs.
- Flag duplicate source URLs for manual review. Shared URLs are not merged automatically because they may represent versions or related records.
- Document corrections and re-run syntax, metadata, and link checks before deployment.
4. Reproducibility
A seed is not sufficient on its own. Reproduction of a synthetic dataset requires the matching engine version, configuration, seed, dependency versions, numerical environment, and any external inputs. Byte-identical output is not claimed across different environments unless a specific validation record demonstrates it.
Companion scenario, validation, episode, strategy, or manifest files support audit of a published run. Their presence does not replace independent review of the model, assumptions, or suitability for a research question.
5. Known limitations
- Some legacy records do not yet have a recorded licence or publisher citation.
- Descriptions and previews may be incomplete or older than the publisher's current version.
- External links can move, require authentication, or become unavailable.
- Catalog inclusion does not establish quality, legal suitability, ethical suitability, or fitness for a particular model.
Corrections
To report a wrong source, licence, citation, version, or description, identify the public record URL and provide a publisher page or other primary evidence.
info@iotdataset.com