evidencelib

evidencelib is a Python library for belief-function calculations in Dempster-Shafer theory (DST) and Dezert-Smarandache theory (DSmT). It provides a compact, zero-dependency quantitative core for reasoning under uncertainty using belief functions.

• Documentation: https://evidencelib.readthedocs.io
• Source Code: https://github.com/itaprac/evidencelib
• PyPI: https://pypi.org/project/evidencelib/
• Issue Tracker: https://github.com/itaprac/evidencelib/issues

Features

• Multiple models — Shafer's classical DST, free DSmT, and constrained hybrid DSm models
• Proposition algebra — symbolic construction of unions, intersections, and parsing from strings
• Fusion rules — Dempster, Yager, Smets/TBM, disjunctive, Dubois-Prade, Hybrid DSm, PCR1–PCR6, PCR5+, PCR6+
• Source handling — Shafer reliability discounting, Jousselme distance
• Belief measures — mass, belief, plausibility, commonality, conflict
• Decision support — generalized pignistic and DSmP probabilities for arbitrary propositions, singletons, and disjoint Venn regions; decisions by BetP, DSmP, maximum belief, or maximum plausibility
• Import/export — round-trip JSON and CSV (exact=True for lossless restoring of computed results), plus publication-ready LaTeX tables
• Optional plotting — mass assignment bars, source comparison heatmaps, Venn region diagrams, and decision plots
• Zero dependencies — pure Python, no external packages required
• Fully typed — type hints throughout the codebase

Installation

pip install evidencelib

For plotting support:

pip install "evidencelib[plot]"

Requires Python 3.10 or later.

Quick Start

Basic DST Example

from evidencelib import Frame

# Create a frame with mutually exclusive hypotheses
frame = Frame.dst(["Alive", "Dead"])
Alive, Dead = frame.symbols()

# Define a basic belief assignment
m = frame.mass({
    Alive: 0.2,
    Dead: 0.5,
    Alive | Dead: 0.3,
})

# Query belief measures
print(m.belief(Alive))        # 0.2
print(m.plausibility(Alive))  # 0.5
print(m.pignistic())          # {"Alive": 0.35, "Dead": 0.65}
print(m.decision())           # "Dead"

Combining Evidence

from evidencelib import Frame

frame = Frame.dst(["A", "B"])
A, B = frame.symbols()

m1 = frame.mass({A: 0.6, A | B: 0.4})
m2 = frame.mass({B: 0.3, A | B: 0.7})

# Dempster's rule
print(m1.dempster(m2).to_dict())
# {"A": 0.512, "A|B": 0.341, "B": 0.146}

# PCR6
print(m1.pcr6(m2).to_dict())
# {"A": 0.54, "A|B": 0.28, "B": 0.18}

DSmT with Overlapping Hypotheses

from evidencelib import Frame

# Free DSm model — hypotheses may overlap
frame = Frame.dsmt(["A", "B"])
A, B = frame.symbols()

m = frame.mass({
    A: 0.2,
    B: 0.3,
    A & B: 0.4,
    A | B: 0.1,
})

print(m.pignistic())          # singleton scores (may not sum to 1)
print(m.pignistic_of(A & B))  # generalized BetP for any proposition
print(m.pignistic_regions())  # probability over disjoint Venn regions

Dynamic Hybrid DSmT

When constraints are learned after source masses were elicited, keep the sources on their original frame and pass the constrained target model during fusion. This preserves the original focal propositions required by the full DSmH S1 + S2 + S3 transfer.

source = Frame.dst(["A", "B", "C"])
A, B, C = source.symbols()
m1 = source.mass({A: 0.1, B: 0.4, C: 0.2, A | B: 0.3})
m2 = source.mass({A: 0.5, B: 0.1, C: 0.3, A | B: 0.1})

# New information establishes that C is non-existent.
target = Frame.hybrid(["A", "B", "C"], exclusive=True, empty=["C"])
combined = m1.dsmh(m2, model=target)
print(combined.to_dict())  # {"A": 0.34, "A|B": 0.41, "B": 0.25}

API Overview

Frames

Propositions

Belief Measures

Fusion Rules

Decision Support

Import and Export

Plotting

Plotting is optional and requires evidencelib[plot].

m.plot()
m.plot_venn()
m.plot_belief_plausibility()
m.plot_pignistic_decision()
m1.plot_comparison(m2, labels=["sensor", "expert"])

Plots keep bar charts simple by default and use a green heatmap colormap. Colors, legends, annotations, and heatmap colormaps can be customized with Matplotlib-compatible values when needed:

m.plot(colors="tab:blue", annotate=False)
m.plot(style="proposition_types")
m1.plot_comparison(
    m2,
    cmap="viridis",
    annotation_text_colors=("black", "white"),
)
m.plot_pignistic_decision(highlight_decision=False)
m.plot_venn(labels=["H1", "H2"], values="mass")

Function forms are also available:

from evidencelib import plot_mass, plot_mass_comparison, plot_venn

Development

git clone https://github.com/itaprac/evidencelib.git
cd evidencelib
python3.10 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev,docs]"

Run tests:

python -m pytest -q

Run the same quality gates as CI:

python -m ruff check .
python -m mypy
python -m coverage run -m pytest -q
python -m coverage report --fail-under=90

Build documentation:

python -m sphinx -W -b html docs docs/_build/html

Examples

See the examples/ directory for complete scripts:

• basic_dst.py — simple DST frame and belief measures
• rules_dst.py — comparing fusion rules
• zadeh.py — Zadeh's classic counterexample
• dsmt_fusion.py — DSmT evidence fusion
• hybrid_dsmt.py — constrained hybrid DSm model
• plotting.py — optional plotting examples

Citing evidencelib

If the evidencelib library contributes to a scientific publication, please acknowledge it by citing "Śniegowski, S., Świder, A., Shekhovtsov, A., & Sałabun, W. (2026). evidencelib: A python library for evidence modeling and fusion under uncertainty. SoftwareX, 36, 103019."

Or using BibTeX:

@article{sniegowski2026evidencelib,
  title={evidencelib: A python library for evidence modeling and fusion under uncertainty},
  author={{\'S}niegowski, Szymon and {\'S}wider, Adrianna and Shekhovtsov, Andrii and Sa{\l}abun, Wojciech},
  journal={SoftwareX},
  volume={36},
  pages={103019},
  year={2026},
  publisher={Elsevier},
  doi={10.1016/j.softx.2026.103019}
}

DOI: https://doi.org/10.1016/j.softx.2026.103019

References

• Shafer, G. (1976). A Mathematical Theory of Evidence. Princeton University Press.
• Smarandache, F., & Dezert, J. (eds.). Advances and Applications of DSmT for Information Fusion.
• Dezert, J., & Smarandache, F. An Introduction to DSmT.
• Zadeh, L. A. (1986). A simple view of the Dempster-Shafer theory of evidence and its implication for the rule of combination. AI Magazine, 7(2), 85-90.

License

MIT License — see LICENSE for details.

Release notes and migration guidance are maintained in CHANGELOG.md.