Elsevier

Information Sciences

Volume 408, October 2017, Pages 41-69
Information Sciences

Choquet based TOPSIS and TODIM for dynamic and heterogeneous decision making with criteria interaction

https://doi.org/10.1016/j.ins.2017.04.037Get rights and content

Abstract

Decision making problems are becoming increasingly complicated. Consequently, the developed methods to support decision makers are more and more general. Group decision-making is a very common scenario. When a team of decision makers is faced with a problem, it is natural to assume that due to diversity of their backgrounds and experience they may require different ways of providing their evaluations and opinions. A very common, and quite unreasonable, assumption made by decision-making methods that deal with groups of decision makers is to consider that all decision makers agreed on a unique criteria set to evaluate given alternatives. Criteria interaction is another common issue in multi-criteria decision making and it is normally addressed using Choquet integral, which alone brings another serious problem of capacity identification. In addition, the decision-making environments are rarely static. Aiming to address all these issues at once, we propose two generalized approaches based on the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) and the TODIM (an acronym in Portuguese for Interactive and Multi-Criteria Decision Making). The new methods are able to consider groups of decision makers with all their different opinions, heterogeneous types of information, criteria interaction, fuzzy measure identification and dynamic environments. The methods are applied in two examples and two case studies showing their feasibility in solving complex real world problems.

Introduction

In the past few decades, the complexity of decision making problems has grown substantially and different aspects must be considered simultaneously. With the growing complexity of the problems, an increasing effort to develop more resourceful methods to support decision makers in their activities is being made.

The TOPSIS is one of the most known MCDM methods and it is very simple and intuitive. Generally speaking, the idea behind TOPSIS is that the best alternative should be as close as possible to the best possible solution and as far as possible from the worst possible solution. The standard formulation of TOPSIS was only able to deal with crisp numbers. This is a very prohibitive limitation. To require that the experts provide their evaluations as crisp numbers, which have only one parameter, demands a precision that they may not have. Other types of data, such as interval number, fuzzy numbers and intuitionistic fuzzy numbers, are very important tools to help the decision makers to express their knowledge more appropriately. As the data types get more general, the number of parameters to be defined increases. For example, an interval number has two parameters, a trapezoidal fuzzy number has four parameters and an intuitionistic trapezoidal fuzzy number has 10 parameters. Obviously, an excessive number of parameter may be problematic. However, it is worth noting that this increase in the number of parameter is, as a matter of fact, an attempt to aid the decision makers to express their knowledge by providing space for uncertainty and ambiguity. Naturally, due to the seriousness of this limitation, the TOPSIS has been extensively extended. There are many variants of TOPSIS to deal with interval number [11], fuzzy information [5], intuitionistic fuzzy data [4], interval-valued intuitionistic fuzzy data [67], probability distributions [39], [61] and hesitant fuzzy [65].

On the other hand, Kahneman and Tversky [24] studied how individuals make decisions in situations involving risks and detected strong bias in some situations, such as loss aversion. In this same work, the Prospect Theory to describe such behavior was proposed. Later, Gomes and Lima [14] proposed one of the first MCDM methods, named TODIM, to present a value function with the same shape of the Prospect function which has been successfully applied in real world problems [16]. However, similarly to TOPSIS, in its standard formulation the TODIM method is only able to process crisp numbers. Naturally, several variants of TODIM have been proposed to deal with fuzzy data [33], heterogeneous data types [13] and hesitant fuzzy [72].

There are still other issues related to MCDM problems. It is becoming increasingly common to have groups of individuals participating in the process of making decisions. The methods to support decision making need to adapt in order to process the input from all stakeholders. Several variants of TOPSIS [36] and TODIM [37] are able to process group decision making. Although one could argue that as long as there is a variant of the method for a single decision maker, one could still solve the group decision making problem by using aggregation operators [64], however, there are some concerns involved that, in fact, also affect most of the variants for group decision making. One major concern is that all decision makers agree with a fixed criteria set and use the same type of data.

There may be situations where it is convenient to work with different types of data in different criteria. For example, in a supplier selection problem while evaluating the price of the product the company may naturally use a crisp number, since the price is precisely provided, but while evaluating the delivery time of the product the company may want to use a random variable. This situation can even be intensified when there are groups of decision makers involved, where different decision makers may want to use different types of data to evaluate an alternative according to a specific criterion. This requires the MCDM methods to be able to process heterogeneous data types. However, although there are many adaptation of both methods, TOPSIS and TODIM, for dealing with several types of data, little effort was made to adapt them for a computation with heterogeneous data. Peng et al. [45] proposed a TOPSIS version for group decision making that deals with crisp numbers, interval values and linguistic variables with different granularity. The TOPSIS was also extended to deal with crisp, interval, fuzzy numbers and Gaussian random variables in [57]. A TOPSIS based method which is able to process different data types was proposed by Li et al. [36]. On the TODIM side, Fan et al. [13] proposed a variant of TODIM to deal with crisp number, interval number and fuzzy data. Of course, there are also other approaches to deal with heterogeneous data types [12].

Although some efforts have been made to process heterogeneous data types, all the above cited methods have some serious shortcomings. The methods in [13], [45], [57] and Espinilla et al. [12] all require transformations of the data types. So, while they accept heterogeneous input, the processing is made with a homogeneous data type. As argued in [41], transformation of the information types may cause some problems.

Another shortcoming of the above mentioned methods is that not all of them can process group decision making and the ones that can require an agreement to be reached on the criteria set. Different decision makers may have different backgrounds and may consider that different criteria are relevant to the problem. Constraining the decision makers to consider a unique criteria set reduces drastically the usefulness of group decision making. Additionally, none of the methods, except in [57], are able to process random variable and even the method in [57] is only able to process Gaussian distributions by transforming them into interval numbers. Random variables are a key type of data since they are a very consolidate method to model uncertainty caused by stochasticity. In what may be considered an even more serious limitation, all the mentioned methods assume a static environment.

Assuming a static environment may lead to an unrealistic modeling. The performance of the alternatives may depend on several factors that can change or are not completely predictable beforehand. For instance, investment profiles can behave differently with the variation of the interest rate. The price of a seasonal product can vary depending on the period of the year. The weather conditions can affect the performance of different strategies in cases of accidents in the sea. Not only that, but also the alternatives can be affected differently by such underlying changes. Interesting works have been done considering the case of heterogeneous preferences relations and dynamic environment [47]. Methods that use preference relations are different of methods based on ratings, such as the TOPSIS and TODIM. For more about heterogeneous preferences relations the interested reader is referred to Ureña et al. [55].

Another interesting issue in MCDM problems is criteria interaction. When individual decision makers define their set of criteria, two or more criteria can present redundant or complementary information. Actually, the hypothesis of criteria independence is rarely verified in practice [20]. Two very important tools to deal with criteria interaction are the fuzzy measure [53] and the Choquet integral [10]. The fuzzy measure relaxes the additivity requirement of the classical measure to the monotonicity requirement. The Choquet integral is based on the fuzzy measure and generalizes a wide range of aggregation operators [17, and references therein]. Naturally, the Choquet integral spread in the MCDM field [19], [46] and several variants of TOPSIS and TODIM that incorporate the Choquet integral have been proposed [4], [15], [71]. However, to apply the Choquet integral it is necessary a fuzzy measure and, with the exponential growth of the number of parameters in a fuzzy measure, this can become a difficult computational problem to solve. Several methods have been proposed for identification of a fuzzy measure [21], [30], [59] but as it will be pointed out in Section 3.5 they have some shortcomings when applied in the MCDM context.

The paper is organized as follows. In Section 2, a quick review of the necessary concepts is included. In Section 3 we introduce the GMC-RTOPSIS and the GMC-RTODIM, a new approach for fuzzy measure identification which is more suitable for MCDM problems and two examples. Section 4 includes two case studies illustrating complex problems in imperfect settings. Finally, Section 5 presents some conclusions and directions for future research.

Section snippets

Basic concepts and definitions

In this section we present the definitions of the data types that will be used in this paper, which are: crisp, interval, fuzzy and intuitionistic fuzzy numbers. Here, the crisp numbers are denoted by lower case letters (a), interval numbers are denoted by bold lower case letters (a), fuzzy sets are denoted by lower case letters with tilde (a˜) and intuitionistic fuzzy sets are denoted by capital letters with tilde (A˜). Also, we present some basic definitions on fuzzy measures and Choquet

Group Modular Choquet (GMC) approaches

In this section, we present the Group Modular Choquet Random TOPSIS (GMC-RTOPSIS) and the Group Modular Choquet Random TODIM (GMC-RTODIM) that handle several practical problems of multi-criteria decision making. The methods are able to process groups of decision makers, different types of information, which here are crisp numbers, interval numbers, fuzzy numbers, intuitionistic fuzzy numbers and random variables. In addition, the methods are able to process underlying factors that are not a

Case Study 1

This problem evaluates the air quality of the city of Guangzhou/China in November of 2006, 2007 and 2008. The air is monitored by sixteen stations across the Pearl River Delta region. From the 16 stations, three were selected to provide the data. The stations measure the concentration of respirable suspended particulate (PM10), sulphur dioxide (SO2) and nitrogen dioxide (NO2). In this situation, the stations would be considered as the decision makers, the months of November of 2006, 2007 and

Conclusion

In this work we introduced new methods for group decision making based on TOPSIS and TODIM, named GMC-RTOPSIS and GMC-RTODIM, respectively. These methods allow each decision maker to act independently if desired and it is able to process several types of information, to name crisp, interval number, fuzzy number, intuitionistic fuzzy number and probability distributions. The methods are able to deal with interacting criteria by incorporating the Choquet integral in the aggregation step. As

Acknowledgment

R. Lourenzutti would like to thank the Brazilian agency CAPES for the financial support and R. A. Krohling would like to thank the financial support of the Brazilian agency CNPq under grant nr. 309161/2015-0.

References (72)

  • L.F.A.M. Gomes et al.

    Criteria interactions in multiple criteria decision aiding: a Choquet formulation for the TODIM method

    Procedia Comput. Sci.

    (2013)
  • M. Grabisch

    The application of fuzzy integrals in multicriteria decision making

    Eur. J. Oper. Res.

    (1996)
  • M. Grabisch et al.

    A review of methods for capacity identification in Choquet integral based multi-attribute utility theory: applications of the kappalab R package

    Eur. J. Oper. Res.

    (2008)
  • W. Jia et al.

    Using neural networks to determine Sugeno measures by statistics

    Neural Netw.

    (1997)
  • I. Kojadinovic

    Modeling interaction phenomena using fuzzy measures: on the notions of interaction and independence

    Fuzzy Sets Syst.

    (2003)
  • I. Kojadinovic

    Estimation of the weights of interacting criteria from the set of profiles by means of information-theoretic functionals

    Eur. J. Oper. Res.

    (2004)
  • I. Kojadinovic

    Minimum variance capacity identification

    Eur. J. Oper. Res.

    (2007)
  • R.A. Krohling et al.

    Fuzzy TOPSIS for group decision making: a case study for accidents with oil spill in the sea

    Expert Syst. Appl.

    (2011)
  • R.A. Krohling et al.

    Combining prospect theory and fuzzy numbers to multi-criteria decision making

    Expert Syst. Appl.

    (2012)
  • P.J.M. van Laarhoven et al.

    A fuzzy extension of Saaty’s priority theory

    Fuzzy Sets Syst.

    (1983)
  • K.-M. Lee et al.

    Identification of λ-fuzzy measure by genetic algorithms

    Fuzzy Sets Syst.

    (1995)
  • D.-F. Li et al.

    A systematic approach to heterogeneous multiattribute group decision making

    Comput. Ind. Eng.

    (2010)
  • R. Lourenzutti et al.

    A study of TODIM in an intuitionistic fuzzy and random environment

    Expert Syst. Appl.

    (2013)
  • R. Lourenzutti et al.

    The Hellinger distance in multicriteria decision making: An illustration to the TOPSIS and TODIM methods

    Expert Syst. Appl.

    (2014)
  • R. Lourenzutti et al.

    TODIM based method to process heterogeneous information

    Procedia Comput. Sci.

    (2015)
  • R. Lourenzutti et al.

    A generalized TOPSIS method for group decision making with heterogeneous information in a dynamic environment

    Inf. Sci.

    (2016)
  • J.-L. Marichal

    Entropy of discrete Choquet capacities

    Eur. J. Oper. Res.

    (2002)
  • J.-L. Marichal et al.

    Determination of weights of interacting criteria from a reference set

    Eur. J. Oper. Res.

    (2000)
  • T. Murofushi et al.

    Non-monotonic fuzzy measures and the Choquet integral

    Fuzzy Sets Syst.

    (1994)
  • H.V. Rowley et al.

    A practical approach for estimating weights of interacting criteria from profile sets

    Fuzzy Sets Syst.

    (2015)
  • R.-C. Tsaur

    Decision risk analysis for an interval TOPSIS method

    Appl. Math. Comput.

    (2011)
  • R. Ureña et al.

    Managing incomplete preference relations in decision making: a review and future trends

    Inf. Sci.

    (2015)
  • S.-P. Wan

    Power average operators of trapezoidal intuitionistic fuzzy numbers and application to multi-attribute group decision making

    Appl. Math. Model.

    (2013)
  • X.-Z. Wang et al.

    Particle swarm optimization for determining fuzzy measures from data

    Inf. Sci.

    (2011)
  • Y.-M. Wang et al.

    A two-stage logarithmic goal programming method for generating weights from interval comparison matrices

    Fuzzy Sets Syst.

    (2005)
  • S. Wierzchoń

    An algorithm for identification of fuzzy measure

    Fuzzy Sets Syst.

    (1983)
  • Cited by (0)

    1

    On leave - Department of Electrical and Computer Engineering, University of Alberta Edmonton, Alberta T6G 2V4, Canada.

    View full text