opendea-allys 0.1.1

Análise Envoltória de Dados (DEA) simples, robusta e extensível em Python.

opendea

Data Envelopment Analysis (DEA) in Python — simple, robust, and extensible, built on SciPy.
Includes CCR (CRS) and BCC (VRS) models, super-efficiency, additive model,
as well as modules for NDEA (two-stage networks) and Dynamic DEA with carry-overs.

⚠️ Note: The SBM (Slack-Based Measure, Tone 2001) is included in the code but inactive in version 0.1.0.
It will be available in a future release (0.2.0).

---

📦 Installation

User (pip)

pip install opendea          # core only
pip install opendea[viz]     # core + plotting (matplotlib, seaborn)
pip install opendea[full]    # everything: plotting + notebooks + dev tools

Development

git clone https://github.com/yourusername/opendea.git
cd opendea
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev,viz]"

---

⚡ Quickstart

import pandas as pd
from opendea import dea_ccr_input, dea_bcc_output
from opendea.plotting import plot_efficiency

df = pd.DataFrame({
    "x1": [4, 2, 3, 5],
    "x2": [2, 1, 1, 3],
    "y1": [1, 1, 1, 2],
}, index=["A","B","C","D"])

# CCR (CRS) input-oriented
res_ccr_in = dea_ccr_input(df, inputs=["x1","x2"], outputs=["y1"])
print(res_ccr_in[["efficiency"]])

# BCC (VRS) output-oriented
res_bcc_out = dea_bcc_output(df, inputs=["x1","x2"], outputs=["y1"])
print(res_bcc_out[["phi"]])

# Plot efficiencies
plot_efficiency(res_ccr_in, title="CCR Efficiency")

---

🧰 Main API

Classical models

• dea_ccr_input(df, inputs, outputs)
• dea_bcc_input(df, inputs, outputs)
• dea_ccr_output(df, inputs, outputs)
• dea_bcc_output(df, inputs, outputs)

Extensions

• super_eff_ccr_input(df, inputs, outputs)
• super_eff_ccr_output(df, inputs, outputs)
• dea_additive_ccr(df, inputs, outputs)
• dea_additive_bcc(df, inputs, outputs)
• dea_sbm_input(df, inputs, outputs, vrs=True) 🚫 inactive in v0.1.0

Advanced

• ndea_two_stage_input(df, inputs_stage1, link_m, outputs_stage2, vrs=True)
• dynamic_dea_input(panels, inputs, outputs, carryovers, vrs=True)

Utilities

• projections(df, inputs, outputs, result, orientation="input"|"output")
• peers_from_lambdas(result)

---

🧠 Conventions (summary)

• Input-oriented: minimize θ
  Projections: x* = θ·x0 − s− ; y* = y0 + s+
• Output-oriented: maximize φ
  Projections: x* = x0 − s− ; y* = φ·y0 + s+
• SBM (ρ): 0–1, average proportional reduction in inputs.
• Results return a DataFrame (or DEAResult in typed API) with columns:
  • efficiency (θ) or phi (φ) or rho (SBM)
  • lambda_* (intensities)
  • s_minus_*, s_plus_* (slacks)

---

🔬 Advanced examples

NDEA (two-stage in series)

from opendea import ndea_two_stage_input
df_net = pd.DataFrame({
  "x1":[4,2,3,5], "x2":[2,1,1,3],
  "m1":[3,2,2,4],              # link Stage1->Stage2
  "y1":[1,1,1,2],
}, index=list("ABCD"))
res_net = ndea_two_stage_input(df_net, ["x1","x2"], ["m1"], ["y1"], vrs=True)
print(res_net[["efficiency"]])

Dynamic DEA with carry-overs

from opendea import dynamic_dea_input
panels = {
  1: pd.DataFrame({"x1":[5,3,4], "y1":[1,1,2], "k1":[2,1,1]}, index=["A","B","C"]),
  2: pd.DataFrame({"x1":[4,3,3], "y1":[2,1,2], "k1":[2,1,1]}, index=["A","B","C"]),
}
dyn = dynamic_dea_input(panels, inputs=["x1"], outputs=["y1"], carryovers=["k1"], vrs=True)
for t, df_t in dyn.items():
    print(t, df_t[["efficiency"]])

---

🗺️ Roadmap

• Cross-efficiency (benevolent/aggressive)
• Window analysis (sliding windows)
• Malmquist TFP
• Multiplier (dual) models and Assurance Region I/II
• Extended NDEA / Dynamic (network-SBM, dynamic-SBM)
• Directional distance functions (DDF), robust/stochastic DEA, sensitivity analysis

---

🧪 Tests

pytest -q

---

🤝 Contributing

• PRs are welcome!
• Run ruff + black before submitting.
• Always add tests.

---

📄 License

MIT — see LICENSE.