bulanik

Easy-to-use fuzzy numbers, fuzzy inference and fuzzy multi-criteria decision making (MCDM) in pure Python, with no runtime dependencies. Bulanık is Turkish for fuzzy. Optional bridges connect it to LumiX (optimization modelling) and sezgi (metaheuristics), and an optional plotting module draws the results in a light and a dark theme.

Documentation: https://tdelphi1981.github.io/bulanik/

Install

pip install bulanik                      # core, no runtime dependencies
pip install "bulanik[lumix]"             # + the LumiX bridge (bulanik.lumix)
pip install "bulanik[sezgi]"             # + the sezgi bridge (bulanik.sezgi)
pip install "bulanik[plot]"              # + figures (bulanik.plot, needs matplotlib)
pip install "bulanik[lumix,sezgi,plot]"  # everything

Python 3.10 to 3.13. For development, clone the repository and run uv sync (see Contributing).

Quickstart

from bulanik import IFN, TFN
from bulanik.aggregate import wa

a, b = IFN(0.6, 0.3), IFN(0.5, 0.4)
print(a + b)                          # IFN(mu=0.8, nu=0.12)
print(a.add(b, ops="einstein"))       # the Einstein operator family
print(wa([a, b], [0.7, 0.3]))         # weighted average
print((TFN(1, 2, 3) + TFN(3, 4, 5)).score())   # crisp score for ranking

Every operator family, score and method is chosen by name (ops="einstein", score("normalized")); an unknown name raises a ValueError that lists the valid ones.

What is in it

• Fuzzy numbers: triangular and trapezoidal; intuitionistic, Pythagorean, Fermatean and q-rung orthopair; spherical, picture and neutrosophic; their interval-valued versions (IVIFN, IVPFN, IVFFN, IVQROFN, IVNeutrosophicFN) and interval type-2 trapezoidal numbers (IT2TrFN). Immutable and validated.
• Aggregation, sets and scales: weighted average and geometric mean, expert aggregation, membership functions, linguistic variables and linguistic scales.
• Fuzzy inference: Mamdani, Sugeno and Tsukamoto systems from plain-text rules; interval type-2 systems with KM, EKM, Nie-Tan and BMM type reduction; ANFIS.
• Decision making: AHP, fuzzy AHP (also over any number kind), SWARA, entropy, CRITIC and PIPRECIA weights; DEMATEL with CFCS; best-worst consistency; TOPSIS, VIKOR, WASPAS, EDAS, CODAS, COPRAS, MARCOS, CoCoSo, WISP and ELECTRE, on crisp matrices or fuzzy ratings (fuzzy CODAS, WASPAS and CoCoSo included), and Chen's fuzzy TOPSIS. Interval-valued AHP (ahp_interval_valued) with the multiplicative and possibility methods.
• Bridges: with LumiX, the linear best-worst method, Zimmermann fuzzy programming and MCDM scores as optimization data; with sezgi, nonlinear and fuzzy BWM and ANFIS training.
• Figures: bulanik.plot for numbers, scales, inference, rankings and DEMATEL.

The documentation has a user guide, an API reference, and 24 runnable examples (also in examples/).

Development

uv sync
uv run python -m pytest -q

The dev group installs LumiX, OR-Tools, sezgi and matplotlib, so the bridge tests and every example run; without an extra, its tests and examples are skipped. Coverage is measured on the pure-Python core.

Citing

Please cite bulanik using CITATION.cff (GitHub's "Cite this repository"), and the original papers behind the methods you use, listed with DOIs in the references.

License

MIT. See LICENSE.