Fuzzy mathematical programming. Methods and applications (Q1310803): Difference between revisions

This book is a misleadingly titled rather confuse presentation of a collection of applications of linear programming techniques to solving optimization problems involving imprecise data. The authors' \textit{Preface} promises that ``this study presents systematically the state-of- the-art of fuzzy mathematical programming in both techniques and applications\(\dots\)'' using ``concepts and methods developed for solving practical problems''. In reality, what we receive is a collection of more or less unrelated models of decision making problems involving ``fuzzy data''. If more carefully and less pretentiously written the substantial content of the book would reduce to, say, a 100 pages story which could adequately been titled `\textit{Ingenious use of the simplex method taught by examples}'. In its actual form, the rather long text is a mixture of not always convincing formal representations of decision problems, numerical examples which, sometimes, having nothing to do with the problems and methods they are supposed to illustrate and straightforward speculations. Chapter 1, the `\textit{Introduction}', implicitly states that the methodology the authors plan to use in their approach of decision making problems involving vaguely defined objectives and restrictions is based on the theory of fuzzy sets. As many people who experienced the difficulty of adapting standard mathematical methods and techniques to solving real-world optimization and decision making problems I was convinced by the works of \textit{H.-J. Zimmermann} [see, e.g., Int. J. Gen. Syst. 2, 209-215 (1976; Zbl 0338.90055); Fuzzy Sets Syst. 1, 45-55 (1978; Zbl 0364.90065); Inf. Sci. 36, 29-58 (1985; Zbl 0578.90095)] and others that fuzzy set theoretical thinking can be a worthy conceptual approach toward more ``flexible'' mathematical programming. I feel that it is only a matter of time until a rigorous theoretical framework will be created for that purpose. However, it seems that the authors have a strong opinion of what the fuzzy set theoretical basis of such an approach should look like: ``Fuzzy set theory is a theory of graded concepts (a matter of degree), but not a theory of chance. Therefore, figures and numerical tables are considered paramount in the study of fuzzy set theory. They, unlike confusing mathematical slang, difficult functions, etc., provide the best way to communicate with outsiders. This is an important concept in the new generation of operations research'' (cf. page 1). Of course, this introductory statement was not intended to offend readers' intelligence. In fact, it does not matter if the authors deliver their ``capsule look into all existing fuzzy linear and nonlinear \textit{mathematical} programming problems, methods, and applications'' (cf. page 1; here, as well as in what follows italization of parts of the text is due to the reviewer) in an informal language (of figures and tables) as long as it is intelligibly done. The methods ``incapsulated'' in the figures and tables can be significant even if ``few [of these] methods have been yet tested with real world problems by real decision makers'' (cf. page 10). The problem is that it is quite difficult (if not impossible) to check the trustworthiness and/or applicability of methods and statements given by figures and numerical tables to other situations than those incapsulated in the specific figures and tables. Also, by informal (and not too careful) handling of essentially mathematical concepts one can obtain strange effects as the following: ``When any of \(f(c,x)\) and/or \(g(A,x)\) is a nonlinear function, [the following equation given on page 2 \[ \text{maximize } z=f(c,x)= cx\quad\text{subject to } g(A,x)= Ax\leq b\quad\text{and}\quad x\geq 0,\tag{1.1} \] where \(c\), \(b\) and \(x\) are vectors and \(A\) is a matrix] becomes a nonlinear programming problem. If \(x\) is restricted to be an integer, then Equation (1.1) will become an integer programming problem.'' (cf. p. 3) Chapter 2, titled `\textit{Fuzzy Set Theory}', is supposed to provide the readers with new conceptual (mathematical ?!) tools to compensate the fact that ``precise mathematics are not sufficient to model a complex system'' (cf. page 14). It is an about sixty pages long mixture of imprecise definitions (typical: al. 2.2.1.12), meaningless examples (typical: al. 2.2.2.1), and unsupported speculations (see, for instance, p. 36). Its basic informative content can be found and much better understood from a paper of \textit{L. A. Zadeh} [Inf. Control 8, 338-353 (1965; Zbl 0139.246)] or, if more technical details are desired, from the book of \textit{G. J. Klir} and \textit{T. A. Folger} [`Fuzzy sets, uncertainty, and information'' (1988; Zbl 0675.94025), Chapter 2] salted with a few pages of \textit{A. Kaufmann} and \textit{M. M. Gupta} [`Introduction to fuzzy arithmetic: theory and applications'' (1991; Zbl 0754.26012)]. The only conclusion I was able to draw from this chapter is that, even in a ``study [which] concentrate[s] on the applications and not pure mathematics'' (cf. p. 35) and in which ``the practical meaning and interpretation of membership are considered more important than mathematical terms'' (cf. p. 35) terminological precision, accuracy of information and logical coherence cannot be divorced from semantic and/or heuristic analysis of formal concepts and models without disastrous effects. As illustrated by most of Chapter 2 (typically, Section 2.3.2), such a divorce pushes conceptual conclusions into the realm of meaningless thinking. It determines reality behind formal concepts, models become un-recognizable to practitioners and, by doing so, it cancels the very reasons for which those formal concepts and models where invented. The previous conclusion also applies to Chapter 3 which is the only part of the work under review having something to do with mathematical programming. Chapter 3 is an attempt to summarize a rich literature which basically shows how imaginative use of linear programming techniques can help solving decision making problem involving fuzzy data. Readers of this chapter trying to learn more than they have previously known on that matter will be disappointed: The uninspired presentation (see Section 3.1.2 among others) renders fruitless any effort of understanding the internal logic of the models the authors attempt to discuss. Readers \textit{already familiar with the literature on ``fuzzy programming''} and looking for numerical examples showing how various models in this area could be applied to down-to-earth problems or interested in classroom exercises may be more lucky: Most of the ``examples'' collected in this chapter are ``readable'' and worth analyzing. They give a taste of what ``fuzzy programming'' is all about. Nevertheless, all readers should be careful since the authors are sometimes allowing themselves deviations from mathematical correctness as, for instance, when repeatedly implying on pages 132, 136 and 152 that the ``traditional linear programming problem'' \[ \max z= cx\quad\text{s.t. }(Ax)_ i\leq b_ i,\;\forall i\quad\text{and}\quad x\geq 0\tag{3.32} \] has ``a unique optimal solution'' when ``\(c\), \(A\) and \(b_ i\), \(\forall i\), are precisely given''. Also, I would not advise my students to implement the recipes of ``solving'' optimization problems the authors call ``algorithms'' since they include recommendations like those on page 138 (see also page 151, etc.): ``Step 5. After referring to Table 3.13, the decision maker is then asked for his \textit{subjective goal} \(b_ 0\) and its \textit{tolerance} \(p_ 0\) for solving a symmetric fuzzy linear programming problem. If the decision maker \textit{does not like to give his goal for the fuzzy objective}, go to Step 6. If \(b_ 0\) is given go to Step 8. Step 6. Solve Problem 3 of Equation (3.39). A unique \textit{Werners solution} is then provided. Step 7. \textit{Is the solution of Problem 3 satisfying?} Consider the following cases: \[ \begin{matrix} &\text{IF} &&\text{THEN}\\ 1) & solution\;is\;satisfied &\longrightarrow &\text{print out results and stop}\\ 2) & the\;user\;has\;realized\;his/her\;goal &\longrightarrow & \text{give the goal \(b_ 0\) and go to Step } 8\\ 3) & \text{resource } i, \text{ for some } i \text{ are }idle &\longrightarrow & decrease\;b_ i\;(\text{and }change\;p_ i)\dots\;.\text{''}\end{matrix} \] Of course, my reluctance to recommend this sort of ``algorithms'' to my students is due to the fact that the authors systematically forget to make precise the meaning of terms as those italized in the text above. In Chapter 4 the authors continue their ``vague'', hardly readable, presentation of applications of the simplex method started in Chapter 3. It is not clear to us why the problems discussed in this section are called ``possibilistic linear programming models'' (cf. p. 188). That is probably done in order to oppose them to the ``stochastic programming approaches'' (cf. p. 187) of decision problems of similar nature. Also, the distinction the authors make between the ``fuzzy mathematical programming problems'' discussed in Chapter 3 and the ``possibilistic programming models'' which form the topics of Chapter 4 remains to be clarified. In any case, some of the numerical examples supposed to illustrate the ``possibilistic'' models are worth analyzing for the simple reason that they seem realistic (see, for instance, paragraphs 4.1.2.1a, 4.1.4.1, 4.2.2a, etc.). Again, students should be aware of the authors' careless use of mathematical terminology and theorems. One of the examples in this sense in Section 4.2.1a where terms as ``classical game'', ``fuzzy matrix game'' and ``fuzzy game'' are used with no proper explanations and the passage from Equation (4.91) to (4.92) and (4.93) is made with no appropriate provisions that the values \(v\) and \(w\) should be different from zero [compare to \textit{T. Parthasarathy} and \textit{T. E. S. Raghavan}, `Some topics in two-person games' (1971; Zbl 0225.90049)]. Chapter 5, titled ``Concluding Remarks'' is mostly an attempt to compare the philosophy underlying the fuzzy set theoretical approach of uncertainty with the philosophy underlying its probabilistic treatment. The conclusion, reached by the authors after some fuzzy argumentation, in the end of Section 5.1 (p. 270), is that ``the fuzzy set theory abandons the laws of the excluded middle (or additivity requirements) and contradiction, and thus abandons the standard probability calculus altogether, since it destroys the de Morgan relations. \textit{Studies of fuzzy set theory are not going to block the search for reality nor restrict the development of probability, but are going to impel humanity to develop its full power}''. In the same combative spirit the authors analyze in Section 5.2 ``stochastic versus possibilistic programming''. The readers, who were already told in Chapter 4 that stochastic programming (as a whole!) lacks computational efficiency and flexibility (see page 186) may be finally convinced of these ``facts'' by comparing the difficulty of solving the non-understandable formulation of the ``nonlinear programming problem'' in Equation (5.3) with the ``easiness'' of solving the completely different and unrelated problem given in Equation (5.4). The authors claim (with no detectable argument) that the latter reduces to a (seemingly unrelated) ``deterministic model'' in Equation (5.7) which turns out to be (big wonder!) ``a linear programming problem'' (see p. 274). After reading this book I ask myself whether it contains at all parts or sections which my students could profit from. And I am sorry to say that the only positive part in this sense is the rich \textit{Bibliography} which may lead them to the original sources of ``fuzzy mathematical programming''.