PROMETHEE II
AvailablePreference Ranking Organization METHod for Enrichment of Evaluations II. Brans et al. (1986)
The idea
PROMETHEE I incomparable pairs: Model 1 ↔ Model 6
Each laptop model is compared pair by pair across criteria: leaving flow φ⁺ (how much it beats others) minus entering flow φ⁻ (how much others beat it) gives the net flow φ.
Show as table
| Alternative | φ⁺ | φ⁻ | φ |
|---|---|---|---|
| Model 4 | 0.6093 | 0.0503 | +0.5591 |
| Model 3 | 0.2687 | 0.0853 | +0.1834 |
| Model 5 | 0.2606 | 0.0893 | +0.1713 |
| Model 2 | 0.1219 | 0.2220 | -0.1002 |
| Model 6 | 0.0365 | 0.3928 | -0.3562 |
| Model 1 | 0.0415 | 0.4989 | -0.4574 |
Data: Laptop selection (Goswami 2020).
PROMETHEE compares alternatives pair by pair on each criterion with a chosen preference function and ranks them by net flow.
When to use it
- You can name negligible (q) and decisive (p) differences.
- Very different scales, and no normalization should decide their spread.
- You want PROMETHEE I to expose incomparable pairs.
Algorithm
1Pairwise differences
Only differences are used, so there is no normalization step and negative values are fine.
2Preference function per criterion
Six types: I usual (any difference counts), II U-shape (q), III V-shape (p), IV level (q, p), V linear with indifference (q, p), VI Gaussian (s).
3Example: type V, linear with indifference
Differences up to q are ignored, differences above p count fully, and preference grows linearly in between.
4Aggregated preference index
5Leaving and entering flows
m is the number of alternatives; the 1986 paper reports plain sums without 1/(m-1), which gives the same ranking.
6Net flow
PROMETHEE II ranks by decreasing net flow, while PROMETHEE I uses the two flows separately and shows pairs where they disagree as incomparable.
Used together with
Why these pairs
- AHP then PROMETHEE (1,484 OpenAlex co-mentions): Dağdeviren (2008), equipment selection. The first AHP-PROMETHEE paper is dated 1995 by Çebi et al. (2022).
- Fuzzy ANP or fuzzy AHP then PROMETHEE: Tuzkaya et al. (2010), material handling equipment.
- Entropy or CRITIC then PROMETHEE is common, but PROMETHEE does not normalize: weights computed on a normalized matrix and thresholds in raw units describe different things.
- As a cross-check next to TOPSIS, VIKOR and COPRAS: Sałabun, Wątróbski & Shekhovtsov (2020).
Reference example
Details
Brans, J.-P.; Vincke, Ph.; Mareschal, B. (1986). How to select and how to rank projects: The PROMETHEE method. European Journal of Operational Research 24(2), 228-238. doi:10.1016/0377-2217(86)90044-5
Tested against the published hydroelectric power station example (six projects, six criteria, one preference type per criterion; input in Mareschal 2019, Table 1) and the Goswami (2020) laptop selection problem (Tables 13-18). The ranking a5 > a2 > a4 > a6 > a3 > a1 matches; pi differs by at most 0.001 and the summed net flow by 0.002 because the 1986 table rounds pi to 3 decimals before summing. Goswami (2020) matches to 0.00001, and pyDecision agrees to 1e-16.
Checked by this site's test suite.
Sources
Show 11 sources
- Brans et al. (1986). How to select and how to rank projects: The PROMETHEE method. European Journal of Operational Research 24(2), 228-238. doi:10.1016/0377-2217(86)90044-5First presented by Brans in 1982 (Université Laval). Companion paper: Brans & Vincke (1985), Management Science 31(6), 647-656.
- Brans, Vincke & Mareschal (1986), European Journal of Operational Research 24(2), 228-238 doi:10.1016/0377-2217(86)90044-5
- Brans & Vincke (1985), Management Science 31(6), 647-656 doi:10.1287/mnsc.31.6.647
- Brans & Mareschal (2005), PROMETHEE methods doi:10.1007/0-387-23081-5_5
- Mareschal (2015), Note on the PROMETHEE net flow computation bertrand.mareschal.web.ulb.be/PG2024/assets/promethee_net_flow.pdf
- Mareschal (2019), Exercises with Visual PROMETHEE bertrand.mareschal.web.ulb.be/PG2024/assets/exercises-vp-2019.pdf
- Goswami (2020), Foundations of Management 12(1), 93-110 doi:10.2478/fman-2020-0008
- Behzadian et al. (2010), 217 PROMETHEE papers reviewed, EJOR 200(1), 198-215 doi:10.1016/j.ejor.2009.01.021
- De Keyser & Peeters (1996), EJOR 89, 457-461 doi:10.1016/0377-2217(94)00307-6
- Verly & De Smet (2013), IJMCDM 3(4), 325-345 doi:10.1504/IJMCDM.2013.056781
- Sałabun, Wątróbski & Shekhovtsov (2020), Symmetry 12(9), 1549 doi:10.3390/sym12091549