Licenses & attribution

PaMIR distributes no data. pamir.download(id) fetches each dataset from its original source into your local cache. You download it directly from that source and are bound by its license and terms — below. The PaMIR package code is Apache-2.0; that covers the code only, never the datasets.

Note on Kaggle licenses. A license shown on a Kaggle dataset is almost always set by the re-uploader, not the rights-holder — a “CC0” there is the uploader’s statement about their upload, and does not transfer rights in the underlying data. Since PaMIR does not redistribute the tables, the practical risk is low; the column below states what each source declares and what it obliges you to do.

Results (AUCs, tables, figures) may be published for all datasets; the obligations concern redistributing the tables themselves and attribution.

Per-dataset terms

dataset	license (as declared by the downloaded source)	your obligation	origin
bankruptcy	Data files © Original Authors (Kaggle)	Primary source is UCI 572 (CC-BY-4.0) — attribution to Liang et al. (2016) required.	source
bondora	CC0 1.0 (Kaggle reuploader)	Set by the reuploader, not Bondora; underlying data is Bondora AS’s public loan book — attribute Bondora AS.	source
conorsully	CC0 1.0	No restrictions; synthetic dataset.	source
dish	CC0 1.0 (Kaggle reuploader)	No stated restrictions; provenance unverified (Kaggle title: “Automobile Loan Default Dataset”).	source
gastonstat	No license (GitHub, no LICENSE file)	No rights granted — research use only; do not redistribute.	source
gmsc	Unknown (Kaggle reuploader)	Primary source is the 2011 Kaggle competition — competition rules; do not redistribute.	source
laotse	CC0 1.0 (Kaggle reuploader)	No stated restrictions.	source
lc_clean	No license (HuggingFace repo, none stated)	LendingClub data via a third party — no rights granted; research use only.	source
lc_my	Unknown (Kaggle reuploader)	No rights granted — research use only.	source
lc_small	ODbL (Open Database License)	Share-alike + attribution: a derived database must also be ODbL.	source
lt_vehicle	Other (specified in description)	L&T hackathon archive — no rights granted; research use only.	source
pakdd	No license (competition archive)	PAKDD 2010 (NeuroTech / UFPE) — research use only.	source
poland_1yr	CC-BY-4.0 (UCI)	Attribution to Zięba et al. (2016) required.	source
poland_3yr	CC-BY-4.0 (UCI)	Attribution to Zięba et al. (2016) required.	source
poland_5yr	CC-BY-4.0 (UCI)	Attribution to Zięba et al. (2016) required.	source
prosper	CC0 1.0 (Kaggle reuploader)	Set by the reuploader, not Prosper; underlying is Prosper Marketplace data.	source
sba	CC0 1.0 (Kaggle reuploader)	Educational subset — cite Li, Mickel & Taylor (2018).	source
south_german	CC-BY-4.0 (UCI)	Attribution to Grömping (2019) required.	source
taiwan	CC0 1.0 (Kaggle “UCI ML” reuploader)	Primary source is UCI 350 (CC-BY-4.0) — attribute Yeh & Lien (2009).	source

Attribution is required by license for bankruptcy, poland_1yr, poland_3yr, poland_5yr, south_german, taiwan (CC-BY-4.0 via their UCI primary source) and lc_small (ODbL share-alike).

Attribution & citations

Cite the source of each dataset you use — pamir.dataset_info(id)["citation"]:

• bankruptcy — Liang, D., Lu, C.-C., Tsai, C.-F., & Shih, G.-A. (2016). Financial ratios and corporate governance indicators in bankruptcy prediction: A comprehensive study. European Journal of Operational Research, 252(2), 561–572.

• bondora — Bondora AS public loan book (Estonia); CC-BY-SA-4.0 — share-alike + attribution

• conorsully — Conor Sully, synthetic credit-score dataset (Kaggle conorsully1/credit-score), MIT

• gmsc — Kaggle “Give Me Some Credit” competition (2011), Credit Fusion & Will Cukierski.

• lc_small — LendingClub via Kaggle (itssuru/loan-data); ODbL — share-alike + attribution required

• pakdd — PAKDD 2010 Data Mining Competition (NeuroTech Ltd. / UFPE). Also benchmarked in Lessmann, Baesens, Seow & Thomas (2015), EJOR 247(1), 124–136.

• poland_1yr — Zięba, M., Tomczak, S. K., & Tomczak, J. M. (2016). Ensemble boosted trees with synthetic features generation in application to bankruptcy prediction. Expert Systems with Applications, 58, 93–101.

• poland_3yr — Zięba, M., Tomczak, S. K., & Tomczak, J. M. (2016). Ensemble boosted trees with synthetic features generation in application to bankruptcy prediction. Expert Systems with Applications, 58, 93–101.

• poland_5yr — Zięba, M., Tomczak, S. K., & Tomczak, J. M. (2016). Ensemble boosted trees with synthetic features generation in application to bankruptcy prediction. Expert Systems with Applications, 58, 93–101.

• sba — Li, M., Mickel, A., & Taylor, S. (2018). “Should This Loan be Approved or Denied?”: A Large Dataset with Class Assignment Guidelines. Journal of Statistics Education, 26(1), 55–66.

• south_german — Grömping, U. (2019). South German Credit Data: Correcting a Widely Used Data Set. Report 4/2019, Beuth University of Applied Sciences Berlin.

• taiwan — Yeh, I.-C., & Lien, C.-H. (2009). The comparisons of data mining techniques for the predictive accuracy of probability of default of credit card clients. Expert Systems with Applications, 36(2), 2473–2480.