Data, Code & Reproducibility Policy

AIDSAE publishes applied AI and data science research that is credible, auditable, and reproducible in principle. We recognise that some datasets (especially health or sensitive geospatial data) cannot be fully open; however, authors must provide enough detail for reviewers and readers to understand exactly what was done and how results were validated.

1) Minimum Reproducibility Requirements (all submissions)

All AIDSAE manuscripts must include:

  • Dataset statement: source, time period, inclusion/exclusion rules, key variables, and any access restrictions or permissions.
  • Preprocessing pipeline: cleaning, feature engineering, missing-data handling, normalization/encoding steps.
  • Validation strategy: train/validation/test split logic or cross-validation design, and how hyperparameters were tuned.
  • Baselines: at least one appropriate baseline comparator (e.g., logistic regression, random forest, standard benchmark, or prior published approach).
  • Metrics: clearly defined metrics (and why they fit the problem).
  • Leakage safeguards: how data leakage was prevented (especially for time-series, patient-level, farm-level, or spatial data).
  • Error analysis: brief but concrete failure modes (where the model performs poorly and why).
  • Reproducibility statement: one of the options below.

2) Reproducibility Statement (choose one option)

Authors must include one of the following:

  • Open code + open data: repository link + dataset link (preferred).
  • Open code + restricted data: repository link + instructions for qualified access to the dataset.
  • Restricted code/data: if code cannot be shared, authors must provide sufficiently detailed methods (model parameters, training procedure, software versions, and pseudocode where appropriate) so the work is reproducible in principle.

3) Model and Reporting Transparency (recommended)

Where applicable, authors should provide:

  • a short Model Card (intended use, limitations, bias risks),
  • computational resources used (hardware, training time),
  • versioned dependencies (major library versions).

4) Sensitive Data (health, minors, location privacy, proprietary datasets)

For sensitive datasets, authors must state:

  • ethics approval or legal basis where applicable,
  • privacy safeguards (de-identification, aggregation),
  • what can/cannot be shared and why.

5) Non-compliance

Manuscripts missing the minimum reproducibility requirements may be returned before peer review or desk rejected.