Decision Support
The 2-rank additive model with axiomatic design in multiple attribute decision making

https://doi.org/10.1016/j.ejor.2020.04.011Get rights and content

Highlights

  • We present five axioms to characterize an ideal 2-rank additive MADM method.

  • We present two 2-rank MADM methods and show their desired properties.

  • We design simulation experiments to justify our proposal.

Abstract

In this paper, we study the 2-rank multiple attribute decision making (MADM) problem where alternatives are divided into two preference-ordered categories. We begin our study by introducing five axioms that characterize an “ideal” 2-rank additive MADM method, and then present an impossibility theorem: Five axioms cannot be satisfied simultaneously due to the intrinsic conflict among them. Next, we propose two simple 2-rank MADM methods: Permutation-based 2-rank method (PRM) and Ranking-based 2-rank method (RRM), and explore their desired properties based on the axioms. Furthermore, we design detailed simulation experiments and find out that the PRM and the RRM share the same 2-rank result in most cases although these two methods violate different axioms. Finally, we illustrate the PRM and the RRM through a consumer preference analysis case study.

Introduction

Multiple attribute decision making (MADM) usually aims to obtain the ranking of alternatives or select the most preferred one(s) involving multiple attributes (Bana e Costa & Oliveira, 2012; Durbach & Stewart, 2012; Greco et al., 2016; Hwang & Yoon, 2012; Wallenius et al., 2008). The techniques of MADM have been widely used in many areas, including engineering, management, economy, and military (Geldermann et al., 2009; Hwang & Yoon, 2012; Ishizaka & Nemery, 2013).

However, in many real-world MADM problems, obtaining a complete ranking of alternatives can be very time-consuming and sometimes unnecessary. This is especially true with a large number of alternatives (Morente-Molinera et al., 2017; Morente-Molinera et al., 2018). Consequently, it is sometimes preferred to classify alternatives into several pre-defined and preference-ordered categories, leading to the multiple criteria/attribute sorting (MCS) problem (Greco et al., 1998; 2002; Pawlak & Słowiński, 1994; Roy, 1996; Zopounidis & Doumpos, 1999). To date, the MCS problem has become a frequently investigated topic (Zopounidis & Doumpos, 2002). Various MCS methods have been proposed in the existing studies. These methods can be classified into four main streams: the value function-based methods (Doumpos et al., 2001; Jacquet-Lagreze & Siskos, 1982; Keeney & Raiffa, 1976), the outranking relation-based methods (Corrente et al., 2013; Liu et al., 2015; Mousseau et al., 2000; Rocha & Dias, 2008), the distance-based methods (Chen et al., 2008; 2011; Vetschera et al., 2010) and the decision rule-based methods (Chakhar & Saad, 2012; Greco et al., 2001; Greco et al., 2010; Kadziński et al., 2014). Generally, thresholds are required in these MCS methods to identify the category that an alternative belongs to.

Although the MADM and the MCS problems have been intensively investigated, two gaps still remain unfilled.

  • (1)

    Typically, decision makers only need to divide alternatives into two ordered categories in practical MCS problems. For example, when evaluating scientific research projects, the science foundation committee has to divide scientific proposals into two categories: the supported and the rejected. Though of great importance, this special case of the MCS has not been sufficiently investigated based on the axiomatic design. To our knowledge, Bouyssou and Marchant (2007) used conjoint measurement to analyze the twofold partitions of alternatives that obtained using “non-compensatory sorting models’’. Also, the twofold MCS problem is related to the 2-rank problem which is presented by Hochbaum and Levin (2006), inspired by Kemeny and Snell's model for the group-ranking problem (Kemeny & Snell, 1962).

  • (2)

    Generally, the threshold of the existing classification models is provided exogenously by the decision maker, who either specifies a threshold in terms of evaluations of what means a “good” alternative, or in terms of cardinality of the set of “good” alternatives (e.g., what are the three best candidates in a job selection problem) (Jacquet-Lagreze & Siskos, 2001; Kadziński & Tervonen, 2013; Mousseau et al., 2001). However, in many actual problems, it is hard for decision makers to provide such thresholds. For instance, a company investigates consumer preference to make marketing plans. The company wants to know what products consumers prefer and what they do not. In this case, the company cannot present the threshold of consumers “preferred” and “not preferred” in advance.

In order to overcome these two shortcomings, we call the twofold MCS the 2-rank MADM in this paper, and use the axiomatic approach to study the 2-rank MADM problem in the framework of multiple attribute additive model, one of the classical MADM models (Keeney & Raiffa, 1976; Von Winterfeldt & Edwards, 1986). Furthermore, differently from the existing classification methods, in our model the threshold is determined endogenously by minimizing the weighted sum of violations. In a sense, we build the 2-rank model that minimizes the assignment conflicts, measured by the weighted sum of “advantages” of the alternatives in the bottom category with respect to the alternatives in the top category. This proposal is carried out by the following points:

  • We present five axioms: Anonymity (ANO), Rankings Invariance (RI), Minimum Intensity of Reversed Rankings (MIRR), Rational Scale Invariance (RSI), and Inversion (INV) to fully characterize the desired properties of an “ideal” 2-rank MADM method. Based on these axioms, we propose an impossibility theorem, which states that all five axioms cannot be satisfied simultaneously due to the intrinsic conflict between RI and MIRR.

  • We develop two simple 2-rank MADM methods: Permutation based 2-rank method (PRM) and ranking based 2-rank method (RRM). We prove that the PRM satisfies ANO, MIRR, RSI, and INV, but violates RI. In contrast, the RRM satisfies ANO, RI, RSI, and INV, but violates MIRR.

  • We design simulation experiments to further explore the properties of the PRM and the RRM, and illustrate the two methods through a consumer preference analysis case study.

The rest of the paper is organized as follows. In Section 2, we briefly introduce the basic knowledge of MADM and define the 2-rank MADM problem. After that, we present five axioms with respect to the 2-rank MADM and the impossibility theorem in Section 3. In Section 4, we propose the PRM and the RRM, and examine the desired properties of them based on the established axioms. Afterwards, we present simulation experiments and a case study to further investigate the properties of the PRM and the RRM in Section 5. Finally, we end this paper with conclusions and future prospective researches in Section 6.

Section snippets

Preliminaries

In this section, we introduce some basic knowledge of MADM, and then define the 2-rank MADM problem.

Axioms in the 2-rank MADM

In this section, we present five axioms that identify an “ideal” 2-rank MADM method. After that, we propose an impossibility theorem, which states that no 2-rank MADM method satisfies all five axioms simultaneously.

Definition 3 (Anonymity (ANO))

Let P=[pij]n×m be an MAEM. Let δ1:{1,2,,n}{1,2,,n} be a permutation on the set of alternatives and δ2:{1,2,,m}{1,2,,m} be a permutation on the set of attributes. In accordance with the permutation of attributes δ2, W is permuted as W¯. Let P¯=[p¯ij]n×m be the MAEM obtained from

Two 2-rank MADM methods

In this section, we propose two simple 2-rank MADM methods: Permutation-based 2-rank method (PRM) and Ranking-based 2-rank method (RRM), and then discuss their properties based on the proposed axioms.

Simulation analysis and an illustrative example

In this section, we conduct two simulation experiments to explore the properties of the PRM and the RRM. After that, we use a consumer preference analysis case to illustrate these two methods.

Conclusions

We investigate the 2-rank problem with axiomatic design in the context of additive MADM. The main contributions are concluded as follows.

  • (1)

    We propose five axioms, ANO, RI, MIRR, RSI, and INV to thoroughly characterize an “ideal” 2-rank MADM method. Based on the axioms, we present an impossibility theorem, which states that no 2-rank MADM method satisfies all five axioms simultaneously due to the conflict between RI and MIRR.

  • (2)

    We present two simple 2-rank MADM methods: The PRM and the RRM. We prove

Acknowledgments

This work was supported by the grants (Nos. 71871149 and 71571125) from NSF of China, and the grants (Nos. sksyl201705, sksyl201821, and 2018hhs-58) from Sichuan University.

References (48)

  • M. Grabisch

    The application of fuzzy integrals in multicriteria decision making

    European Journal of Operational Research

    (1996)
  • S. Greco et al.

    Rough sets theory for multicriteria decision analysis

    European Journal of Operational Research

    (2001)
  • S. Greco et al.

    Rough sets methodology for sorting problems in presence of multiple attributes and criteria

    European Journal of Operational Research

    (2002)
  • E. Jacquet-Lagreze et al.

    Assessing a set of additive utility functions for multicriteria decision-making, the UTA method

    European Journal of Operational Research

    (1982)
  • E. Jacquet-Lagreze et al.

    Preference disaggregation: 20 years of MCDA experience

    European Journal of Operational Research

    (2001)
  • M. Kadziński et al.

    Stochastic ordinal regression for multiple criteria sorting problems

    Decision Support Systems

    (2013)
  • M. Kadziński et al.

    Robust ordinal regression for dominance-based rough set approach to multiple criteria sorting

    Information Sciences

    (2014)
  • J. Liu et al.

    A group decision-making approach based on evidential reasoning for multiple criteria sorting problem with uncertainty

    European Journal of Operational Research

    (2015)
  • J.A. Morente-Molinera et al.

    Solving multi-criteria group decision making problems under environments with a high number of alternatives using fuzzy ontologies and multi-granular linguistic modelling methods

    Knowledge-Based Systems

    (2017)
  • J.A. Morente-Molinera et al.

    A group decision making support system for the Web: How to work in environments with a high number of participants and alternatives

    Applied Soft Computing

    (2018)
  • V. Mousseau et al.

    Using assignment examples to infer weights for ELECTRE TRI method: Some experimental results

    European Journal of Operational Research

    (2001)
  • V. Mousseau et al.

    A user-oriented implementation of the ELECTRE-TRI method integrating preference elicitation support

    Computers & Operations Research

    (2000)
  • Z. Pawlak et al.

    Rough set approach to multi-attribute decision analysis

    European Journal of Operational Research

    (1994)
  • R. Vetschera et al.

    Robustness and information levels in case-based multiple criteria sorting

    European Journal of Operational Research

    (2010)
  • Cited by (19)

    • Exploring 2-rank strategic weight manipulation in multiple attribute decision making and its applications in project review and university ranking

      2023, Engineering Applications of Artificial Intelligence
      Citation Excerpt :

      For example, Pekelman and Sen (1974) and Yager (2019, 2002) studied how to defend against strategic manipulation from the viewpoint of aggregation approaches in group decision making (GDM). In comparison, Dong et al. (2021a), Wu et al. (2020), and Zhang et al. (2018a) investigated how to detect and manage a series of noncooperative behaviors in GDM consensus reaching processes from different perspectives. Furthermore, Dong et al. (2022), Mathew et al. (2020), and Xu et al. (2021) developed different management mechanisms to manage noncooperative behaviors in large-scale group consensus reaching processes.

    • Data-driven method to learning personalized individual semantics to support linguistic multi-attribute decision making

      2022, Omega (United Kingdom)
      Citation Excerpt :

      Multi-attribute decision-making (MADM) involves an individual or a group of decision makers selecting an option by evaluating a set of alternatives according to multiple attributes [1,2,3,4].

    • Classification-based strategic weight manipulation in multiple attribute decision making

      2022, Expert Systems with Applications
      Citation Excerpt :

      For instance, Hochbaum and Levin (2016) presented the concept of the 2-rank problem in group decision making (GDM), a special classification-based problem, and developed a simple approximation algorithm to solve the 2-rank problem. Wu, Wu, Dong, Liang, and Zhao (2020) proposed two simple 2-rank MADM methods, including the permutation and ranking-based 2-rank methods, and provided some relevant theoretical analysis. Kadziński et al. (2021) proposed a preference disaggregation method for value-based multi-decision sorting problems.

    View all citing articles on Scopus
    View full text