Sensitivity analysis in MCDM using the weight space (Q1200790)

The solution by computer of the TRIMAP (three-objectives linear program) method is presented. The method applies to problems of the following form: \[ \text{Max} f(x)=Cx\quad\text{ s.t. }x \in F=\{x\in\mathbb{R}^ n:Ax=b,\;x\geq 0\} \tag{1} \] where \(C=[c_ 1,c_ 2,c_ 3]\), \(c_ k\in\mathbb{R}^ n\) for \(k=1,2,3\). To help the decision maker in ``learning'' the shape of the nondominated surface in the interactive procedure an indifference region is attached to each solution. After that a sensitivity analysis is made in the case of a perturbation of the objective function matrix depending on a scalar parameter: \(C(y)=C+yD\), where \(D=[d_ 1,d_ 2,d_ 3]\) is a constant matrix and \(y\in\mathbb{R}\). In that case an indifference region \(y\in[y_{min},y_{max}]\) is determined and this result is used thereafter to approaching the indifference region for \(y\in\mathbb{R}^ 3\). A more complicated problem arises when a perturbation of the right-hand side is considered: \(b(w)=b+wh\), \(w\in\mathbb{R}\). In this case there is no need to discuss an indifference region for an efficient point, since that point can be an unfeasible point. In that case the authors use a method which is similar to the corresponding primal dual potential analysis in the \(MC^ 2\)-simplex method, introduced by Seiford and Yu. After this, briefly, the cases are studied, when problem (1) is modified by the introduction of a new constraint or a new decision variable. The paper is closed by an illustrative example, which contains copies of the computer screen during the execution of the TRIMAP method.