Elsevier

Applied Soft Computing

Volume 12, Issue 8, August 2012, Pages 2178-2187
Applied Soft Computing

Satisficing solutions of multi-objective fuzzy optimization problems using genetic algorithm

https://doi.org/10.1016/j.asoc.2012.03.002Get rights and content

Abstract

In the present paper, a genetic algorithm for multi-objective optimization problems with max-product fuzzy relation equations as constraints is presented. Since the non-empty feasible domain of such problems is, in general, a non-convex set; the traditional optimization methods cannot be applied. Here, we are presenting a genetic algorithm (GA) to find “Pareto optimal solutions” for solving such problems observing the role of non-convexity of the feasible domain of decision problem. Solutions are kept within feasible region during the mutation as well as crossover operations. Test problems are developed to evaluate the performance of the proposed algorithm and to determine satisficing decisions. In case of two objectives, weighting method is also applied to find the locus of optimal solutions.

Highlights

► A multi-objective optimization problem with max-product fuzzy relation equations as constraints is presented. ► Pareto optimal solutions are obtained for solving such problems observing the role of non-convexity of the feasible domain of decision problem. ► Solutions are kept within feasible region during mutation and crossover operations. ► Test problems are developed for evaluating performance and determining satisficing decisions. ► In case of two objectives, weighting method is also applied to find the locus of optimal solutions.

Section snippets

Background and motivation

Max–min fuzzy relation equations were first studied by Sanchez [21] in 1976. Later on, Pedrycz [19] studied the max-product fuzzy relation equations. Different types of fuzzy relation equations have been studied since then [2], [4], [8], [13], [17], [20], [23]. Linear optimization problem with fuzzy relation equations as constraints was first considered by Fang and Li [5] with max–min composition and by Loetamonphong and Fang [14] with max-product composition. For both the compositions,

The problem

Consider the following system of fuzzy relation equations:xA=b,where A = [aij]m×n, 0  aij  1, is a m × n dimensional fuzzy matrix, b = [b1, b2, …, bn], 0  bj  1, is a n-dimensional vector, and “∘ ” denotes the max-product composition of x and A. Let I = {1, 2, …, m} and J = {1, 2, …, n} be the index sets. Given fuzzy relation matrix A, output vector b, and relation (1), input vectors x = [x1, x2, …, xm] are decidable.

Let X(A, b) = {x  [0, 1]m|x  A = b} be the feasible domain of fuzzy relation Eq. (1). For any x1, x2  X,

Analysis and formation of genetic algorithm

A genetic algorithm (or GA) is a search technique used in computing to find true or approximate solutions to optimization and search problems. The scalar concept of “optimality” does not apply directly in the multi-objective setting. A useful replacement is the notion of Pareto optimality. In our case, the solution space is, in general, a non-convex set and the objective functions considered are not necessarily linear. Therefore, the only one consideration in deriving the satisficing solutions

Randomly generating max-product fuzzy relation equations

To obtain the measurement of performance of proposed genetic algorithm, test problems are considered by taking multi-objective objective functions and constructing a feasible problem by randomly generating a fuzzy matrix A and constructing vector b from A according to some criteria. The procedure of generating feasible fuzzy relation equations is as follows:

1.Generate a m × n matrix A whose elements are random numbers from [0, 1].
2.Generate vector b such that its jth element is a random number

Summary of the proposed genetic algorithm

The summary of the proposed method can be summarized as under:

  • 1.

    Initialize maximum number of generations, Gen = 1000.

  • 2.

    Set the current generation number c = 1.

  • 3.

    Generate initial population using Algorithm 1.

  • 4.

    Perform selection procedure and establish the efficient set ES by putting all efficient solutions of the initial population into it.

  • 5.

    Perform mutation procedure using Algorithm 3.

  • 6.

    Perform crossover operation using Algorithm 4.

  • 7.

    Again apply selection/reproduction procedure and update the efficient set ES.

  • 8.

Linear weighting method

In case of multi-objective optimization, most practical problems require the simultaneous multiple objectives. In applications of optimization techniques, one way of obtaining the solution to such problems is by combining the objectives into a single one according to some utility function. To solve the multi objective problems, the linear weighting method is used. The linear weighting method [3], combines all the objectives into a single scalar parameterized objective function by using

Illustrations

In this section, implementation of the proposed genetic algorithm for multiple linear and nonlinear objective functions is discussed. We consider multi-objective linear and nonlinear optimization problems and randomly generated fuzzy relation equations with max-product composition to investigate the nature of the Pareto optimal solution set. Some of the test problems are from well known sources [15].

Conclusion

This paper discusses a multi-objective optimization problem with max-product fuzzy relation equations. The problem is simplified by finding out the fixed components for all the solutions. Genetic algorithm is used to derive the satisficing decisions of the problem by finding Pareto optimal solutions. The feasibility of solutions is always maintained during the crossover and mutation operations. The efficient set is constantly updated and finally hill climbing method is employed to further

Acknowledgements

Authors are thankful to CSIR for supporting this work under CSIR grant no. 25(0167)/08/EMR-II. Authors are grateful to all the reviewers for their valuable suggestions and comments.

References (25)

  • A.V. Markovskii

    On the relation between equations with max-product composition and the covering problem

    Fuzzy Sets and Systems

    (2005)
  • W. Pedrycz

    On generalized fuzzy relational equations and their applications

    Journal of Mathematical Analysis and Applications

    (1985)
  • Cited by (24)

    • Solving max-Archimedean t-norm interval-valued fuzzy relation equations

      2022, Fuzzy Sets and Systems
      Citation Excerpt :

      They proposed a genetic algorithm to find the Pareto optimal solutions. More work on multi-objective optimization problem subject to max-t FRE and max-product FRE can be seen in Guu et al. [7] and Thapar et al. [36] respectively. When a system is modelled as a system of FRE, then sometimes a transition from FRE to interval-valued FRE becomes crucial.

    • An application of OWA operators in fuzzy business diagnosis

      2017, Applied Soft Computing Journal
      Citation Excerpt :

      Since Zadeh [16], there have been developments in modelling uncertainty, describing human behavior and complex systems distinguished by incomplete information and multiple subjective variables. Fuzzy prediction models are an alternative in this context ([17–29], among others) for overcoming many of the limitations of the traditional models (selection of the dependent variable, selection of the sample, choice of the explanatory variables, and the treatment and inclusion of the sorted errors).2 Vigier and Terceño [17] and Scarlat [18], among others, have advantages over other prediction models on the basis of diagnostic theory, which establishes the process of business failure and the formal relationship between causes and symptoms.

    • Matrix elementary transformations in solving systems of fuzzy relation equations

      2015, Applied Soft Computing Journal
      Citation Excerpt :

      In 1976, Sanchez [24] first introduced the theory of systems of fuzzy relation equations with sun-inf composition over complete Brouwerian lattices. Since then, the theory has been applied in many fields, such as fuzzy control [17], discrete dynamic systems [10], intelligence technology [18], fuzzy inference systems [11], coding/decoding images and videos [13], and in many other fields [5,7,9,28]. Many works about solving systems of fuzzy relation equations were published [3,8,16,19,21–23,25–27,29,30,32].

    • A new algorithm for resolution of the quadratic programming problem with fuzzy relation inequality constraints

      2014, Computers and Industrial Engineering
      Citation Excerpt :

      Also, a necessary condition for its optimal solution in terms of the maximum solution of FRE was presented by Guu and Wu (2002). We can find other generalizations on these kinds of problems with FRE constraints, for example, see references (Chang & Shieh, 2013; Guu & Wu, 2010; Hassanzadeh, Khorram, Mahdavi, & Mahdavi-Amiri, 2011; Lu & Fang, 2001; Thapar, Pandey, & Gaur, 2012; Zhou & Ahat, 2011). This section is formed by two subsections.

    • A hybrid OC-GA approach for fast and global truss optimization with frequency constraints

      2014, Applied Soft Computing Journal
      Citation Excerpt :

      Since 1980s, evolutionary algorithms (such as genetic algorithm (GA)) and swarm algorithms (such as particle swarm optimization (PSO), artificial bee colony algorithms (ABCA), mine blast algorithm (MBA)) have been greatly developed. These heuristic methods are particularly effective for global and nonlinear optimization problems [5–13] and has been recently applied in structural optimization [13–15]. Among these population-based methods, GA is proposed earlier and more extensively tested by numerical functions and structural benchmark examples.

    • A Genetic Algorithm for Solving Nonlinear Optimization Problem with Max-Archimedean Bipolar Fuzzy Relation Equations

      2023, International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems
    View all citing articles on Scopus
    View full text