DEAPack

DEAPack is a source-audited Python toolkit for data envelopment analysis, efficiency measurement, productivity analysis, and environmental performance. The redesigned 2.x API uses the lowercase deapack import and a common data, technology, result, visualization, and reporting architecture.

2.0.0rc1 is the draft release-candidate identity. The candidate has not yet been published. Its 33 dataset fingerprints now have exact item-level license and attribution mappings; GitHub and PyPI publication actions remain separate manual steps. After those authenticated steps complete, the release candidate will be suitable for public testing and exact version-pinned research; it will still not be the stable 2.0 compatibility promise.

Installation

After publication, install the exact release candidate with:

python -m pip install "DEAPack==2.0.0rc1"

Python 3.10, 3.11, 3.12, and 3.13 are supported. NumPy, pandas, and SciPy are the only required runtime dependencies; ordinary linear programmes use the HiGHS solver bundled through SciPy. Install DEAPack[viz] for the optional Matplotlib result views.

Quick start

import pandas as pd

from deapack import BCCInput, DEAData

frame = pd.DataFrame(
    {
        "dmu": ["A", "B", "C", "D"],
        "input": [1.0, 2.0, 3.0, 4.0],
        "output": [1.0, 3.0, 4.0, 4.0],
    }
)
data = DEAData.from_frame(
    frame,
    dmu="dmu",
    inputs="input",
    outputs="output",
)

result = BCCInput().fit(data)
print(result.summary())
print(result.peers("D"))

Results expose named summary, target, slack, peer, dual, component, and diagnostic tables only when the corresponding numerical and economic account has been certified. They can also produce a self-contained HTML brief and a deterministic audit archive.

Scope

The candidate covers the major classical radial and non-radial families, directional and generalized-distance analysis, price-informed economic efficiency, undesirable-output and environmental technologies, Malmquist and related productivity accounts, network and dynamic production, panel models, and radial metafrontiers. The installed discovery catalog is the authoritative list of public methods:

from deapack import list_methods

for method in list_methods():
    print(method.method_id)

Statistical inference, Färe--Primont productivity, generic congestion, automatic EBM calibration, and several source-incomplete variants remain explicit next-version work rather than provisional APIs.

Documentation and migration

• Package Documentation source
• English Handbook source
• Chinese Handbook catalogs
• 2.x migration guide
• Release notes
• Contribution guide
• Issue tracker

DEAPack 2.x is a greenfield API. Historical import DEAPack and ProdPack scripts require an explicit migration; the 2.x wheel intentionally provides only import deapack.

Citation and license

Use the repository's CITATION.cff and record the exact pre-release version and commit used. The DEAPack software component is licensed under GPL-3.0-only. Bundled dataset content is released only when its provenance record identifies confirmed licensing authority, an approved redistribution status, content license, attribution, and required notice; it does not inherit GPL merely by being represented in a Python file. All 33 current dataset fingerprints have exact mappings: 30 project-created or independently selected fixtures and one external dataset use CC-BY-4.0, while two external datasets retain upstream MIT notices. Package Documentation and Bilingual Handbook Preview 1 are separate components and are cited and licensed separately.