All methods

ELECTRE III

Available

ELimination Et Choix Traduisant la REalité III. Roy (1978)

The idea

Descending distillation, ascending distillation, and net credibility
A7 (3·1)+0.2336A9 (1·3)+0.1911A4 (2·4)-0.0199A10 (4·5)+0.1019A1 (9·1)+0.2991A5 (5·6)-0.1100A3 (5·8)-0.0399A11 (5·8)-0.1032A8 (5·10)-0.0572A2 (10·6)-0.1605A6 (11·11)-0.3349

Incomparable pairs (r↓ vs r↑): A1 ∥ A3, A1 ∥ A4, A1 ∥ A5, A1 ∥ A8, A1 ∥ A9, A1 ∥ A10, A1 ∥ A11, A2 ∥ A3, A2 ∥ A8, A2 ∥ A11, A4 ∥ A7, A7 ∥ A9

Each waste treatment option is compared by fuzzy credibility σ(a, b): descending distillation r↓ selects from the top down, ascending distillation r↑ eliminates from the bottom up, and Φ shows net credibility.

Show as table
Alternativer↓r↑Φ
A731+0.2336
A913+0.1911
A424-0.0199
A1045+0.1019
A191+0.2991
A556-0.1100
A358-0.0399
A1158-0.1032
A8510-0.0572
A2106-0.1605
A61111-0.3349

Data: Waste incineration strategy (Wang and Triantaphyllou 2008).

ELECTRE III checks for every pair whether A outranks B, using indifference (q), preference (p) and veto (v) thresholds.

When to use it

  • Non-compensation matters: environmental, safety or regulatory limits.
  • Imprecise data, and q, p (optionally v) can be stated per criterion.
  • The audience accepts a partial order with incomparable pairs.

Algorithm

  1. 1Advantage of b over a

    Thresholds q (indifference), p (preference) and v (veto) are evaluated at g_j(a) (direct thresholds), with 0 <= q <= p <= v.

  2. 2Partial and global concordance
  3. 3Partial discordance

    Without a veto threshold, d is always 0.

  4. 4Credibility

    If no d exceeds C the product is empty and sigma = C, while any d = 1 gives sigma = 0.

  5. 5Distillation

    Qualification Q(a) is the number of alternatives a outranks minus the number that outrank a; descending distillation keeps the best Q at shrinking cut levels, ascending the worst, and the result is their intersection, where disagreement means incomparable.

Used together with

Why these pairs
  • Subjective weights (AHP, direct rating, stakeholder votes) then ELECTRE III is the usual pattern, because ELECTRE weights are voting powers. Example: Hokkanen & Salminen (1994), solid waste management.
  • SMAA-III for robustness when weights or thresholds are uncertain (Tervonen, Figueira, Lahdelma et al.).
  • As a non-compensatory cross-check next to PROMETHEE II or TOPSIS in the same study; Brans et al. (1986) already compare PROMETHEE with ELECTRE III.

Reference example

Wang, X.; Triantaphyllou, E. (2008)Matches the printed digits
Details

Wang, X.; Triantaphyllou, E. (2008). Ranking irregularities when evaluating alternatives by using some ELECTRE methods. Omega 36(1), 45-63. doi:10.1016/j.omega.2005.12.003

Published example (Section 3.2, pp. 51-53). Waste incineration strategy, 11 alternatives, 11 criteria, direct relative thresholds, no veto. All 110 credibility cells match at 2 decimals, and both distillation pre-orders and the final result (A7 and A9 on top, incomparable) match. In the modified case the top alternative matches; one lower class is a tie where the paper prints a strict order.

Checked by this site's test suite.

Sources

Show 7 sources
  1. Roy (1978). ELECTRE III : un algorithme de classements fondé sur une représentation floue des préférences en présence de critères multiples. Cahiers du Centre d'Études de Recherche Opérationnelle 20(1), 3-24. No DOI. The standard modern description is Figueira, Mousseau & Roy (2005).
  2. Roy (1978), Cahiers du CERO 20(1), 3-24
  3. Figueira, Mousseau & Roy (2005), ELECTRE methods doi:10.1007/0-387-23081-5_4
  4. Wang & Triantaphyllou (2008), Omega 36(1), 45-63 doi:10.1016/j.omega.2005.12.003
  5. Govindan & Jepsen (2016), EJOR 250(1), 1-29 doi:10.1016/j.ejor.2015.07.019
  6. Hokkanen & Salminen (1994), ELECTRE III for solid waste management doi:10.1007/978-94-017-0767-1_9
  7. Tervonen, Figueira, Lahdelma et al., SMAA-III doi:10.1007/978-1-4020-9026-4_15