De novo programming with fuzzy coefficients and multiple fuzzy goals (Q2367736)

The paper, a continuation of an earlier one discussing the same subject [see the authors, ibid. 153, No. 1, 97-111 (1990; Zbl 0719.90092)], presents an approach to the multiple criteria de novo programming which makes use of fuzzy methodology. The de novo multiple criteria programming problem is expressed in the form of a problem with multiple fuzzy goals and with fuzzy coefficients in the objectives and the constraints. Let \(\alpha\) denote the minimal possibility degree of all the coefficients and \(\beta\) the degree to which a solution satisfies the fuzzy goals for a given possibility \(\alpha\). The authors propose two ways of determining such values of \(\alpha\) and \(\beta\), for which the overall satisfaction \(\lambda=\min\{\alpha,\beta\}\) attains its maximal value. One way involves the solution of a linear problem with the parameter \(\alpha\) in the constraints, combined with some search procedure (not indicated in the paper) for different values of \(\alpha\). The other way consists in solving a nonlinear programming problem. Once the desired values of \(\alpha\) and \(\beta\) are found, a second phase of the procedure of solving the de novo programming problem leads to a non-dominated solution with the satisfaction degree \(\lambda=\min\{\alpha,\beta\}\). This solution is found through solving a linear programming problem. Apart from the general case, the authors also consider two special cases: (a) when the membership functions of the coefficients are linear and one- sided; (b) when additionally the decision maker gives the goals and the tolerances for all the objectives. The corresponding special forms of the problems to be solved (those from the general procedure) are derived.