On solutions and distribution problems of the linear programming with fuzzy random variable coefficients (Q1319122)

The authors consider two linear programming models in which the coefficients are so called fuzzy random variables. In the first model this is true only for the objective function coefficients. The other coefficients as well as the decision variables are -- as we can guess on the basis of the notation defined earlier -- usual random variables. In the second model all the coefficients and all the decision variables are fuzzy random variables. Both models are analyzed with the definitive aim of getting a so called fuzzy probability function, projective distribution function and expectation (all these notions are defined in the paper) of the optimal value of the objective function -- which, with the assumptions made, is also a random fuzzy variable. The authors understand a random fuzzy variable as a mapping (function) of the probabilistic universe into the set of fuzzy numbers, with the following property: for each \(\alpha\in (0,1]\), if we take its \(\alpha\)- cut, we receive a random interval or, as the authors state it, a pair of random variables (the end points of a random interval). Unfortunately the paper is very ``heavy'' to read and raises a lot of doubts. The practical value of the algorithm of solving linear problems with random coefficients, proposed in earlier papers of the authors and constituting a basis of the analysis of the initial problems in the present paper, is questionable. What the authors propose is simply to solve the problem for each elementary even \(\omega\in \Omega\) (the problem becomes then a crisp one). This can obviously make sense only in very special cases, i.e. when \(\Omega\) is finite, common for all the coefficients and contains a small number of elementary events. In practice we do not know the probabilistic universe and observe only the values of random variables. It is on the basis of these observations that we determine the frequency function or the distribution function. Apart from that, in such models the random variables are usually not discrete, but rather continuous.