MEREC
AvailableMethod based on the Removal Effects of Criteria. Keshavarz-Ghorabaee et al. (2021)
The idea
Each segment is one alternative's drop in overall performance when the criterion is removed; together they make E.
Show as table
| Criterion | A1 | A2 | A3 | A4 | A5 | Removal effect E | Weight |
|---|---|---|---|---|---|---|---|
| C1 | 0.7394 | 0.0891 | 0.4984 | 0.3818 | 0.0000 | 1.7087 | 0.5752 |
| C2 | 0.0000 | 0.0160 | 0.0032 | 0.0127 | 0.0100 | 0.0419 | 0.0141 |
| C3 | 0.0000 | 0.5198 | 0.0755 | 0.4073 | 0.1902 | 1.1928 | 0.4016 |
| C4 | 0.0130 | 0.0015 | 0.0075 | 0.0000 | 0.0051 | 0.0271 | 0.0091 |
Data: Five alternatives, four criteria (Keshavarz-Ghorabaee et al. 2021).
MEREC removes each criterion in turn and measures how much the alternatives' overall scores change; criteria whose removal changes the scores most get the largest weights.
When to use it
- Weights from the data alone, every value a positive amount.
- A comparison next to CRITIC, Entropy and SD.
Algorithm
1Normalize so that smaller is better
The best value maps to the smallest n; this is the reverse of the WASPAS normalization.
2Overall performance
3Performance without criterion j
The denominator stays n, as in Eq. (4) of the paper.
4Removal effect
S' is never above S, so E is 0 exactly when the column is constant.
5Weights
Used together with
Why these pairs
- MEREC only weighs; published studies rank with CoCoSo, EDAS (Bektaş 2022), MARCOS, WASPAS, CODAS, MABAC (Ecer and Ayçın 2022), WISP (Ulutaş et al. 2022) and COBRA (Popović et al. 2022).
- MEREC next to Entropy, LOPCOW, CRITIC and equal weights, then MARA, RAM and PIV for material selection: Dua et al. (2024); their Case 2 and Case 3 MEREC weights match ours to 4 decimals.
- The origin paper compares MEREC with CRITIC, Entropy and SD (section 4.2, correlations 0.85 to 0.98).
Reference example
Details
Keshavarz-Ghorabaee, M.; Amiri, M.; Zavadskas, E.K.; Turskis, Z.; Antucheviciene, J. (2021). Determination of Objective Weights Using a New Method Based on the Removal Effects of Criteria (MEREC). Symmetry 13(4), 525. doi:10.3390/sym13040525
Illustrative example (Section 4.1, Tables 2-4, 5 x 4): our weights match the printed 4 decimals (largest difference 3.9e-5), and N, S, all 20 values of S' and E match the printed 3 and 2 decimals. The printed E are rounded, so the weights do not follow from them exactly. The comparative example (Section 4.2, Tables 5-6, 10 x 7) matches to 3 decimals, and Dua et al. (2024), Cases 2 and 3, to 4.
Checked by this site's test suite.
Sources
Show 7 sources
- Keshavarz-Ghorabaee et al. (2021). Determination of Objective Weights Using a New Method Based on the Removal Effects of Criteria (MEREC). Symmetry 13(4), 525. doi:10.3390/sym13040525
- Keshavarz-Ghorabaee, Amiri, Zavadskas, Turskis & Antucheviciene (2021), Symmetry 13(4), 525 doi:10.3390/sym13040525
- Dua, Duc, Bao & Trung (2024), EUREKA: Physics and Engineering (2), 131-148 doi:10.21303/2461-4262.2024.003171
- Ulutaş, Stanujkić, Karabašević, Popović & Novaković (2022), Strategic Management 27(4), 23-29 doi:10.5937/straman2200013u
- Popović, Pucar & Smarandache (2022), Journal of Process Management and New Technologies 10(3-4), 66-74 doi:10.5937/jouproman2203066p
- Bektaş (2022), BDDK Bankacılık ve Finansal Piyasalar Dergisi doi:10.46520/bddkdergisi.1178359
- Ecer & Ayçın (2022), Informatica doi:10.15388/22-infor494