Przejdź do treści
Biblioteka skilli
Nauka i badania · Analiza

mcdm

/mcdm

Przeprowadza analizy wielokryterialne metodami AHP, TOPSIS, PROMETHEE i ELECTRE, z testami czułości i odwrócenia rankingu.

Repozytorium
MCDM-Skill ↗
Licencja
MIT
Ostatnia zmiana repozytorium
03.11.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
18
Skrypty
10
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 03.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co sprawdzić przed instalacją
  • 10 skryptów w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.

Skrypty: scripts/mcdm_cli.py, scripts/mcdm_core.py, scripts/mcdm_fuzzy.py, scripts/mcdm_network.py, scripts/mcdm_outranking.py, tests/test_mcdm_core.py, tests/test_mcdm_fuzzy.py, tests/test_mcdm_network.py, tests/test_mcdm_outranking.py, tests/test_pymcdm_crosscheck.py

Co robi i kiedy

Przeprowadza analizy wielokryterialne metodami AHP, TOPSIS, PROMETHEE i ELECTRE, z testami czułości i odwrócenia rankingu.

Oryginalny opis (EN)

Runs and audits multi-criteria decision-making (MCDM/MCDA) studies with tested code: AHP, fuzzy AHP, ANP, DANP (DEMATEL-based ANP), BWM, entropy, CRITIC and ROC weights; TOPSIS, fuzzy TOPSIS, VIKOR, fuzzy VIKOR, PROMETHEE I/II, ELECTRE I/II/III/IV, ELECTRE Tri-B sorting, GRA and WSM/WPM/WASPAS; DEMATEL and fuzzy DEMATEL; weight sensitivity, rank reversal, SMAA-style acceptability and ranking similarity. Use for supplier/site/project/technology selection, resource allocation, criteria weighting, interdependent criteria, sorting into ordered classes, or reviewing an MCDM paper. Triggers: MCDM, M

Jak działa

  1. Modules
  2. Workflow
  3. Running it
  4. Rules that prevent common errors
  5. References in this skill
Sekcje z nagłówków SKILL.md, w oryginale · 90 linii, 659 słów
Podgląd SKILL.md
---
name: mcdm
description: >-
  Runs and audits multi-criteria decision-making (MCDM/MCDA) studies with tested
  code: AHP, fuzzy AHP, ANP, DANP (DEMATEL-based ANP), BWM, entropy, CRITIC and
  ROC weights; TOPSIS, fuzzy TOPSIS, VIKOR, fuzzy VIKOR, PROMETHEE I/II, ELECTRE
  I/II/III/IV, ELECTRE Tri-B sorting, GRA and WSM/WPM/WASPAS; DEMATEL and fuzzy
  DEMATEL; weight sensitivity, rank reversal, SMAA-style acceptability and
  ranking similarity. Use for supplier/site/project/technology selection,
  resource allocation, criteria weighting, interdependent criteria, sorting into
  ordered classes, or reviewing an MCDM paper. Triggers: MCDM, MCDA, AHP, ANP,
  DANP, 層級分析, 網路分析法, 模糊 AHP, TOPSIS, VIKOR, 模糊 VIKOR, PROMETHEE,
  ELECTRE, ELECTRE TRI, BWM, DEMATEL, 灰關聯, GRA, 熵權法, CRITIC, 多準則決策, 權重,
  一致性比率, 敏感度分析. Not for continuous multi-objective optimisation (use
  pymoo) or picking a point on a Pareto front without criteria data.
metadata:
  version: "1.1.0"
  author: "張明瀚 (cmh3601)"
  license: MIT
compatibility: Python >= 3.10 with numpy and scipy; pymcdm and pyDecision optional for cross-checks.
---

# MCDM

Every number in an MCDM result must come from the bundled code or from a
library call, never from mental arithmetic. The implementations in `scripts/`
are tested in `tests/` against pymcdm, pyDecision (where conventions match),
hand-derived cases and crisp limits of the fuzzy methods. `README.md` in this
folder lists every method with its function and call pattern.

## Modules

| Module | Contents |
|---|---|
| `mcdm_core.py` | weights (AHP, fuzzy AHP, BWM, entropy, CRITIC, std, ROC, AIJ/AIP, hybrid); TOPSIS, fuzzy TOPSIS, VIKOR, PROMETHEE I/II, ELECTRE I, WSM, WPM, WASPAS, GRA; DEMATEL; sensitivity, rank reversal, acceptability, rank correlations |
| `mcdm_network.py` | ANP (supermatrix builder, weighted supermatrix, exact Cesàro limit), DANP |
| `mcdm_outranking.py` | ELECTRE II, III, IV, Tri-B; Roy's distillation |
| `mcdm_fuzzy.py` | fuzzy VIKOR, fuzzy DEMATEL, CFCS and other defuzzification, TFN arithmetic |
| `mcdm_cli.py` | runs a whole study from a JSON spec |

## Workflow

1. **Structure the problem.** Goal, alternatives (m), criteria (n), each
   criterion's direction, unit and data source. Check for double counting.
2. **Decide whether criteria are independent.** If not, map the dependence with
   DEMATEL (or fuzzy DEMATEL) and weight with DANP, or build an ANP network.
3. **Choose weights and justify them.** Subjective: AHP, fuzzy AHP, BWM (2n − 3
   comparisons), ROC. Objective: entropy, CRITIC, std. Interdependent: ANP, DANP.
   Hybrid: `combine_weights`. Group: AIJ / AIP.
4. **Check consistency.** AHP CR < 0.10 (Saaty RI for n <= 10, else pass `ri=`);
   inspect `most_inconsistent`. BWM: `xi`, `CR_input`. Fuzzy AHP: modal CR. ANP:
   CR of every pairwise block (`supermatrix_from_comparisons` records them).
5. **Rank (or sort) with at least two families**: distance (TOPSIS, VIKOR, GRA),
   outranking (PROMETHEE, ELECTRE), utility (WSM, WASPAS). For assignment to
   ordered classes use ELECTRE Tri-B and report both pessimistic and optimistic
   results. See `references/methods.md` for choosing.
6. **Test robustness**: `weight_sensitivity`, `rank_reversal_check`,
   `weight_acceptability`, `compare_rankings`.
7. **Report** with `references/reporting-checklist.md`.

## Running it

```bash
python scripts/mcdm_cli.py scripts/example_spec.json --out results.json
python scripts/mcdm_cli.py scripts/example_danp_spec.json
```

Spec keys: `weights.method` in `given | equal | entropy | critic | std | roc |
ahp | fuzzy_ahp | bwm | danp` (+ optional `combine_with`, `alpha`); `methods`
list from `wsm, wpm, waspas, topsis, vikor, promethee, gra`; optional blocks
`electre_i`, `electre_ii`, `electre_iii`, `electre_iv`, `electre_tri`;
`analysis` with `sensitivity`, `rank_reversal`, `acceptability`. Pairwise
entries may be strings like `"1/3"`.

Python use (ANP networks, fuzzy ratings, custom pipelines):

```python
import sys; sys.path.insert(0, "<skill-dir>/scripts")
import mcdm_core as mc, mcdm_network as mn, mcdm_outranking as eo, mcdm_fuzzy as mf
res = mc.topsis(X, w, ["cost", "benefit", "benefit"]); res.ranks  # 1 = best
out = eo.electre_iii(X, w, types, q, p, v); out["descending"]["classes"]
d = mn.danp(Z, groups=[[0, 1, 2], [3, 4]]); d["weights"]
```

Run the tests after any change: `python -m pytest <skill-dir>/tests -q`.

## Rules that prevent common errors

- Ranks are 1 = best and come from `mc.rank`; never `np.argsort(scores) + 1`.
- VIKOR / fuzzy VIKOR Q is "lower is better"; check `Result.higher_is_better`.
- Weights must sum to 1; the code refuses otherwise.
- State the normalisation, thresholds (PROMETHEE q/p/s, ELECTRE q/p/v, c/d
  levels, Tri-B cut λ and profiles), VIKOR v, GRA ρ, TFN scale and
  defuzzification. They are modelling choices that change results.
- ELECTRE III/IV give two pre-orders and a partial final order; incomparability
  (`R`) is a result, not an error. `rank_by_average` is a convenience complete
  order, not Roy's method; say so if you use it.
- ANP limit matrices are computed exactly (spectral projector), so cyclic
  networks converge. For reducible networks pass `reference=` (e.g. the goal).
- Use Buckley's fuzzy AHP, not Chang's extent analysis (Wang, Luo & Hua, 2008).
- Library defaults differ from the textbook in several places (pymcdm TOPSIS
  normalisation, pyDecision Tri-B pessimistic rule, ...): `references/library-map.md`.

## References in this skill

- `README.md` — method catalogue with function names and invocation.
- `references/methods.md` — formulas, method choice, citations.
- `references/fuzzy.md` — TFN arithmetic, scales, fuzzy AHP/TOPSIS/VIKOR/DEMATEL, CFCS.
- `references/library-map.md` — pymcdm / pyDecision equivalents and verified differences.
- `references/reporting-checklist.md` — what a journal reviewer will look for.
Pliki w paczce (18)
  • README.md 8 KB
  • SKILL.md 6 KB
  • references/fuzzy.md 5 KB
  • references/library-map.md 4 KB
  • references/methods.md 14 KB
  • references/reporting-checklist.md 2 KB
  • scripts/example_danp_spec.json 1 KB
  • scripts/example_spec.json 2 KB
  • scripts/mcdm_cli.py 7 KB
  • scripts/mcdm_core.py 30 KB
  • scripts/mcdm_fuzzy.py 5 KB
  • scripts/mcdm_network.py 8 KB
  • scripts/mcdm_outranking.py 13 KB
  • tests/test_mcdm_core.py 11 KB
  • tests/test_mcdm_fuzzy.py 4 KB
  • tests/test_mcdm_network.py 4 KB
  • tests/test_mcdm_outranking.py 7 KB
  • tests/test_pymcdm_crosscheck.py 2 KB

Wersje

W bazie nie ma innych skilli o nazwie mcdm.

Migawka bazy z 13.11.2026 · wersja paczki 882588f913b8 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 96capacity
    Dostarcza wiedzę o ocenie nośności wodnej, liczeniu wskaźników oraz modelu TOPSIS z wagami AHP-CRITIC.
  • 100decision-matrix-mcda
    Szereguje opcje względem ważonych kryteriów macierzą decyzyjną i analizą wielokryterialną (macierz Heuera i Phersona, porównania AHP Saaty'ego z testem spójności) oraz sprawdza odporność zwycięzcy na zmiany wag i ocen.
  • 98xlsx
    Przetwarza surowe dane Excel i CSV dla zadań MCM/ICM: czyści braki i wartości odstające, normalizuje dla TOPSIS, AHP i EWM oraz analizuje korelacje.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.