Interactive fuzzy programming with preference criteria in multiobjective decision-making (Q1814157): Difference between revisions

The following problem is considered: \(\min f(x)\), \(x\in X\subseteq R^ n\), where \(f\) is a vector function having \(k\) components. It is assumed that a set \(U\) of \(r\) \((r\geq k)\) nondominated solutions \(U=\{x^*_ u\), \(u=1,2,\dots,r\}\) has already been selected. The aim is to choose from among them the best one, according to some criteria given by the DM. The values \(pa_ i(f_ i(x^*))\) \((i=1,2,\dots,k)\), expressing the actual percentage of achievement of the objective \(f_ i\) by a certain solution \(x^*\) (evaluated over the entire range of feasible values of \(f_ i\)), are defined. The DM is expected to provide: (1) the values \(pc_ i\) \((i=1,2,\dots,k)\), i.e. the desirable degree of achievement of the optimal value of the objective \(f_ i\) (between 1 and 100), (2) the values \(E_ i\), i.e. the maximal acceptable underachievement size for the objective \(f_ i\), (3) the value \(E_ g\), i.e. the maximal acceptable difference between the actual value of \(\sum^ k_{i=1} pa_ i\) and the value \(PA_ G=\max\sum^ k_{i=1} pa_ i\). The criteria for selecting the best solution are as follows: (1) criteria \(C(i)\) \((i=1,2,\dots,k)\): \(pc_ i-E_ i\leq pa_ i\), (2) criterion \(C(k+1) PA_ G-\sum^ k_{j=1} pa_ j\leq E_ G\). For each criterium \(C(i)\) \((i=1,\dots,k+1)\) a fuzzy set \(U_ i\subset U\) is defined, with the membership function \(h_ i\) having the following properties: (a) \(h_ i(x^*)=0\) for \(x^*\) not satisfying the \(i\)th criterion, (b) \(h_ i(x^*)\) close to 0 for weakly desirable \(x^*\) (i.e. satisfying the \(i\)th criterion not much better than with the equality sign), (c) \(h_ i(x^*)\) close to 1 for strongly desirable \(x^*\) (i.e. \(pa_ i\) close to 100 or --- for \(i=k+1-\sum^ k_{j=1} pa_ j=PA_ G)\). Subsequently, the problem \(\max_{x^*\in U}\min_{1\leq j\leq k+1}\{h_ j(x^*)\}\) is solved. An interactive algorithm is provided, which makes it possible for the DM to vary the values \(pc_ i\), \(E_ i\) \((i=1,\dots,k)\) obtained for different values of \(pc_ i\), \((i=1,\dots,k)\), and \(E_ G\) is proposed.