Data and code package for CERT-FJSP computational-evidence reporting experiments

Description

This dataset provides the public data and code package for the CERT-FJSP computational-evidence reporting study. CERT-FJSP is a solver-agnostic reporting and traceability framework for already-generated flexible job shop scheduler outputs. The package supports artifact-level inspection and selected partial reproduction of manuscript-supporting empirical summaries, including projection-ambiguity metrics, tuple-ablation information-retention outputs, reporting-baseline comparison, reference-policy sensitivity, global-versus-regime masking, and practical interpretive scenario summaries. The archive contains selected Python scripts, supporting source modules, frozen public-facing output artifacts, expected-output documentation, schema and field-glossary material, selection-protocol documentation for the 322 inspected tuple-bearing rows, a 19-dimension retention-scoring rubric, package manifests, verification reports, and provenance-oriented inspection material. It also includes a bounded external public benchmark vignette and an anonymized retrospective industrial handover motivation note. The package is intended for data/code transparency, artifact-level inspection, and selected partial reproduction. It is not a flexible job shop solver, not a benchmark suite, not a solver-selection tool, not a fallback-control package, not regulatory certification, not third-party approval, not independent evidence validation, and not external evidence review. It does not contain manuscript draft files, manuscript source files, cover-letter materials, response-to-reviewers files, raw thesis files, or internal planning materials.

Steps to reproduce

1. Download and extract the archive. 2. Read the root-level README.md, VERIFY.md, SCRIPT_STATUS.md, and PACKAGE_MANIFEST.md files to understand the package scope, included files, runnable entry points, script-status classifications, and claim boundaries. 3. Optionally create a clean Python environment and install the Python dependencies listed in requirements.txt. 4. Run the archive-level verification entry point from the package root: python scripts/verify_archive.py A successful run verifies the public archive structure, JSON validity, checklist references, portability checks, and CSV row-width consistency. 5. Inspect the documentation files under docs/, especially: - public_package_scope.md - schema_and_field_glossary.md - frozen_manuscript_evidence_index.md - expected_output_checklist.md - selection_protocol_322_tuple_rows.md - wp32_retention_scoring_rubric_19_dimensions.md - projection_ambiguity_metrics_report.md 6. Inspect the released outputs under outputs/real_runs/ and, where applicable, compare them with the corresponding expected_outputs/ files and the expected-output checklist. 7. Use the selected scripts under scripts/ as documented inspection or generation-logic artifacts. Only the archive-verification script is intended to run directly from the public package root without omitted upstream inputs; other scripts may require upstream internal artifacts that are outside the public archive scope. 8. Interpret all outputs within the documented claim boundaries. The package supports artifact-level inspection and selected partial reproduction of the reporting, traceability, projection-loss, reference-sensitivity, and regime-masking summaries. It is not a solver implementation, benchmark suite, solver-selection tool, fallback-control package, regulatory certification package, third-party approval package, independent evidence-validation package, or external evidence-review package.

Institutions

• University of Pannonia

  Veszprém, Veszprém

Categories

Licence