Code For Pu-learning
Reproduction materials for a PLOS ONE manuscript on positive-unlabeled (PU) learning for student-aid screening. The package contains the code, the generated datasets, and the numerical results behind the reported tables and figures. All data are simulated; no real student records are included. Datasets: 100 complete baseline cohorts, results/cohorts/baseline/42_data.npz to 141_data.npz. Each file is one generated cohort of 12,000 students with 14 feature columns, true hardship labels, observed aid-record labels, and the exact training/test split (8,400/3,600). DATASETS.md describes the schema and gives a loading example. GitHub repository: https://github.com/yangyangXu7/student-aid-pu-learning
Authors
- Yangyang Xu (ORCID: https://orcid.org/0009-0002-5406-5961)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-16
- DOI
- https://doi.org/10.5281/zenodo.24292754
- Primary Topic
- Machine Learning and Data Classification
- Type
- article
- Field-Weighted Citation Impact
- 0.00