Replication package for "Strategic Cost Efficiency in AI-Driven Financial Management: A Case Study of DeepSeek’s Digital Transformation" (CFRI submission, v1.0)

Replication package for “Strategic Cost Efficiency in AI-Driven Financial Management: A Case Study of DeepSeek’s Digital Transformation” (CFRI submission, v1.0). This dataset contains the exact code, metadata, and analysis-ready files needed to reproduce all figures, tables, diagnostics, and robustness checks reported in Sections 4–6 of the manuscript. All inputs originate from public sources; no proprietary data were used.

Contents

/code/ — scripts organized as 0_fetch/ (download public inputs), 1_clean_transform/ (apply transformations & priors), 2_analysis/ (models, diagnostics, robustness), 3_figures_tables/ (renders all outputs), plus a single entry point run_all.{sh|bat|py|R} and a random-seeds.txt.

/data/processed/ — analysis-ready CSVs used by the paper. (If redistribution of originals is restricted, raw files are not included; use /code/0_fetch/ to retrieve them.)

/metadata/ — provenance_sources.csv mapping each input to Appendix C (C.1–C.2) with URL and access date; transforms_priors.xlsx mirroring Appendix D; variables_dictionary.xlsx (names, units, and construction).

/env/ — environment files (requirements.txt or environment.yml) and optional session info to ensure reproducibility.

/output/ — auto-generated figures, tables, and diagnostic artifacts that match the manuscript.

Reproducibility
Run the repository root script (run_all…) to: (1) fetch public inputs (when raw files are not redistributed), (2) rebuild processed datasets, and (3) regenerate every figure/table for §§4–6, including diagnostics and robustness panels. The package is platform-agnostic; instructions are provided for a standard Python/Conda setup (optional R/Stata scripts are included where relevant).

Licensing & citation
Data are released under CC BY 4.0; code under MIT. Please cite this dataset’s DOI and the associated article. A CITATION.cff file is provided for citation managers.

Notes

The CHANGELOG.md documents versioning (v1.0 = submission). Post-acceptance updates will add the article DOI and any minor alignment fixes.

This record may be embargoed until journal acceptance; metadata remain public.

Corresponding author: Marco I. Bonelli (ORCID 0000-0003-3463-6421).