Abstract
This paper examines the problem of combining a set of ordinal rankings to form an acceptable consensus ranking. The objective of traditional group decision making problem is to determine the Minimum Violation Ranking. Motived by the applications of adjusted consensus in recent years, we study this problem from a new perspective, for obtaining an acceptable consensus ranking for group decision making. In this paper, every voter ranks a set of alternatives respectively, and we know the acceptability index, which represents the minimum adjustments that are allowed for each voter. The problem is to find the Minimum Acceptable Violation Ranking (MAVR) which minimizes the sum of voter’s unacceptable violations. Besides, we develop a branch and bound ranking algorithm to solve this problem. The suggested improvement include: (1) analysing the ranking preference by two ways: pairwise preference and ranking-based preference; (2) constructing the lower bound and upper bound, which exclude at most half of the feasible solutions in each iteration process. Furthermore, the effectiveness and efficiency of this algorithm are verified with an example and numerical experiments. Finally, we discuss two extensions of the basic MAVR problem: the Minimum Weighted Acceptable Violation problem, whose voters are accompanied with a set of weights or multiples, and the Minimum Hierarchy Acceptable Violation problem, which uses hierarchical acceptability indexes. In addition, our results can be applied to other ranking and subset selection problems in which provide consensus rankings over the alternatives.
Similar content being viewed by others
References
Ali I, Cook WD, Kress M (1986) On the minimum violations ranking of a tournament. Manag Sci 32(6):660–672
Arrow KJ (1951) Social choice and individual values, 2nd edn. Wiley, New York
Barzilai J, Cook WD, Kress M (1986) A generalized network formulation of the pairwise comparison consensus ranking model. Manag Sci 32(8):1007–1014
Ben-Arieh D, Chen Z (2006) Linguistic-labels aggregation and consensus measure for autocratic decision making using group recommendations. IEEE Trans Syst Man Cybern Part A Syst Hum 36(3):558–568
Ben-Arieh D, Easton T (2007) Multi-criteria group consensus under linear cost opinion elasticity, vol 43, 713th edn. Elsevier, Amsterdam, pp 721–721
Ben-Arieh D, Easton T, Evans B (2009) Minimum cost consensus with quadratic cost functions. IEEE 39:210–217
Blin JM (1976) A linear assignment formulation of the multiattribute decision problem. Revue française d’automatique, d’informatique et de recherche opérationnelle Recherche opérationnelle 10(2):21–32
Blin JM, Whinston A (1974) Note-a note on majority rule under transitivity constraints. Manag Sci 20(11):1439–1440
Bowman VJ, Colantoni C (1973) Majority rule under transitivity constraints. Manag Sci 19(9):1029–1041
Bustince H, Jurio A, Pradera A, Mesiar R, Beliakov G (2013) Generalization of the weighted voting method using penalty functions constructed via faithful restricted dissimilarity functions. Eur J Oper Res 225(3):472–478
Cook WD, Kress M (1990) A data envelopment model for aggregating preference rankings. INFORMS 36:1302–1310
Cook WD, Saipe A (1976) Committee approach to priority planning: the median ranking method. Cahiers du Centre d’EÉtudes de Recherche Opérationnelle 18:337–351
Cook WD, Seiford LM (1978) Priority ranking and consensus formation, vol 24. INFORMS, pp 1721–1732
Cook WD, Seiford LM (1982) On the borda-kendall consensus method for priority ranking problems. Manag Sci 28(6):621–637
Cook WD, Golan I, Kress M (1988) Heuristics for ranking players in a round robin tournament, vol 15. Elsevier, Amsterdam, pp 135–144
Cook WD, Golany B, Penn M, Raviv T (2007) Creating a consensus ranking of proposals from reviewers partial ordinal rankings, vol 34. Elsevier, Amsterdam, pp 954–965
De Borda JC (1781) Mémoire sur les élections au scrutin. Histoire de l\(\backslash \)’Academie Royale des Sciences
Dong Y, Chen X, Herrera F (2014) Minimizing adjusted simple terms in the consensus reaching process with hesitant linguistic assessments in group decision making. Inf Sci 297(C):95–117
Dong Y, Li CC, Herrera F (2015) An optimization-based approach to adjusting unbalanced linguistic preference relations to obtain a required consistency level. Inf Sci 292(5):27–38
Even S (2011) Graph algorithms. Cambridge University Press, Cambridge
Herrera F, Herrera-Viedma E et al (1996) A model of consensus in group decision making under linguistic assessments. Fuzzy Sets Syst 78(1):73–87
Herrera-Viedma E, Herrera F, Chiclana F (2002) A consensus model for multiperson decision making with different preference structures. IEEE Trans Syst Man Cybern Part A Syst Hum 32(3):394–402
Herrera-Viedma E, Alonso S, Chiclana F, Herrera F (2007) A consensus model for group decision making with incomplete fuzzy preference relations. IEEE Trans Fuzzy Syst 15(5):863–877
Inada KI (1969) The simple majority decision rule. Econometrica 37(3):490–506
Isaak G, Narayan DA (2004) Complete classification of tournaments having a disjoint union of directed paths as a minimum feedback arc set, vol 45. Wiley, New York, pp 28–47
Kemeny JG, Snell L (1962) Preference ranking: an axiomatic approach. Mathematical models in the social sciences. Blaisdell, New York, pp 9–23
Kendall MG (1948) Rank correlation methods
Kendall MG (1955) Further contributions to the theory of paired comparisons. Biometrics 11(1):43–62
Lucas WF (1983) Measuring power in weighted voting systems. Springer, Berlin
Rademaker M, De Baets B (2014) A ranking procedure based on a natural monotonicity constraint, vol 17. Elsevier, Amsterdam
Wei TH (1952) The algebraic foundations of ranking theory. University of Cambridge, Cambridge
Zhang B, Dong Y, Xu Y (2013) Maximum expert consensus models with linear cost function and aggregation operators. Comput Ind Eng 66(1):147–157
Zhang B, Dong Y, Xu Y (2014) Multiple attribute consensus rules with minimum adjustments to support consensus reaching. Knowl-Based Syst 67(3):35–48
Zhang G, Dong Y, Xu Y, Li H (2011) Minimum-cost consensus models under aggregation operators. IEEE Trans Syst ManCybern Part A Syst Hum 41(6):1253–1261
Acknowledgments
This work was partially supported by the NSFC (Grant No. 71601152), and by the China Postdoctoral Science Foundation (Grant No. 2016M592811) and by the Natural Science Basic Research Plan in Shaanxi Province of China (Program No. 2015JM7372).
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: The proof of equivalency of the two linear models
To prove the equivalency, we just need to prove that the two models have the same feasible solution set and the feasible solutions of these two models are the same.
The first linear model is given by
s.t.
The second linear model can be represented as below.
s.t.
where \( A^{l}=(a_{ij}^{l}) \) is the initial preference matrix of voter \( e_{l}, B^{l}=(b_{ij}^{l})\) is the adjusted preference matrix of voter \( e_{l} \) after \( k^{l} \) adjustments \( ( k^{l}\leqslant \alpha ), C=(c_{ij} ) \) is a consensus preference decision matrix.
It is obvious to know that the feasible solutions are the set of all possible rankings, so the feasible solutions of these two models are the same. For example, there are 3 alternatives(or candidates) a, b and c in an election, the feasible solution set can be represented as: \( \{a, b, c\}, \{a, c, b\}, \{b, a, c\}, \{b, c, a\}, \{c, a, b\}, \{c, b, a\}\).
For a given preference decision matrix C (a feasible solution), we will know the distance between each voter \( A^{l} \) and C .
In the first model, because \( d(A^{l},C)= d(A^{l},B^{l})+d(A^{l},C) \), to minimizing the distance between adjusted preference matrix \( B^{l} \) and C , each voter should adjuste their preference as more as possible (at most \( \alpha \)), nearing to C . If the distance between adjusted preference matrix \( A^{l} \) and C is greater than \( \alpha \), \( d_{APD}(A^{l},B^{l}) = \alpha \), then \(d_{APD}(B^{l},C) = d_{APD}(A^{l},C) - \alpha \). If the distance between adjusted preference matrix \( A^{l} \) and C is not greater than \( \alpha \), \( d_{APD}(A^{l},B^{l}) = d_{APD}(A^{l},C) \), then \( d_{APD}(B^{l},C) = 0 \).
In the second model, because \( d(A^{l},C)= d(A^{l},B^{l})+d(A^{l},C) \), to minimizing the adjusted distance between adjusted preference matrix \( B^{l} \) and \( A^{l} \), the distance between \( B^{l} \) and C is as great as it could be (at most \( \alpha \)). If the distance between adjusted preference matrix \( A^{l} \) and C is greater than \( \alpha \), \( d_{APD}(B^{l},C) = \alpha \), then \(d_{APD}(A^{l},B^{l}) = d_{APD}(A^{l},C) - \alpha \). If the distance between adjusted preference matrix \( A^{l} \) and C is not greater than \( \alpha \), \( d_{APD}(B^{l},C) = d_{APD}(A^{l},C) \), then \( d_{APD}(A^{l},B^{l}) = 0 \).
Eventually, we choose the minimum distance among all feasible solutions C as the final decision. Because the objectives of these two linear models are same for every feasible solution, so the final decisions of these two models are the same.
From the above we can draw a conclusion that these two two linear models are equivalent.
Appendix 2: The supplemental proof of the upper bound
Because the objective is to minimize the violations, any one of the feasible solutions is no less than the final solution. To prove the upper bound, we just need to prove that the upper bound we represents, is a feasible solution.
If \( x_{i},x_{j}\in X_{2} \): The worst case is that we choose all the pairs with a higher aggregated violation \( max\left\{ s_{ij},s_{ji}\right\} \) between \( \left( x_{i}\prec x_{j}\right) \) and \( \left( x_{j}\succ x_{i}\right) \). We have \( \sum _{i}\sum _{j} max\left\{ s_{ij},s_{ji}\right\} \ge \sum _{l}\sum _{i}\sum _{j}w_{ij}^{l} \).
Fixing (or choosing) two different pairs of preferences with the highest aggregated violation does not violate the rule of transitivity, they still can be part of some feasible solutions, because the rule of transitivity only exists in more than or equal to three different pairs. For example, \( \left\{ a\succ b, b\succ c\right\} \) is not contradict with the transitivity property, however, \( \left\{ a\succ b, b\succ c, c\succ a \right\} \) is contradict with the rule of transitivity.
Among these alternatives which we have not yet assigned to the prefix order (\( x_{i},x_{j}\in X_{2} \)), we choose two different pairs of preferences with the highest aggregated violation as part of solution. For these two pairs , the violation is \( 2*\left( min^{1}\left\{ s_{ij}\right\} +min^{2}\left\{ s_{ij}\right\} \right) \). For the other pairs, the violation is \( \sum _{i}\sum _{j} max\left\{ s_{ij},s_{ji}\right\} -2*\left( max^{1}\left\{ s_{ij}\right\} +max^{2}\left\{ s_{ij}\right\} \right) \). Because the objective is to minimize the violations, any one of the feasible solutions is no less than the final solution. We have \( \sum _{i}\sum _{j} max\left\{ s_{ij},s_{ji}\right\} -2*\left( max^{1}\left\{ s_{ij}\right\} +max^{2}\left\{ s_{ij}\right\} \right) +2*\left( min^{1}\left\{ s_{ij}\right\} +min^{2}\left\{ s_{ij}\right\} \right) \ge \sum _{l}\sum _{i}\sum _{j}w_{ij}^{l} \).
From the above and the previous proof in Lemma 2 we can draw a conclusion that \( M_{0}+M_{2}\) is an upper bound.
Rights and permissions
About this article
Cite this article
Luo, K., Xu, Y., Zhang, B. et al. Creating an acceptable consensus ranking for group decision making. J Comb Optim 36, 307–328 (2018). https://doi.org/10.1007/s10878-016-0086-9
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10878-016-0086-9