Abstract:

The sensitivity analysis method of multiple criteria decision making is mainly constructed by weight perturbation. When weight assignment is difficult or disputes occur, the evaluation results are often not robust enough. Partial order sensitivity analysis takes decision function as objective function and weight space as constraint condition to establish programming problem, and constructs the partial order relationship between schemes with the help of extreme point set. Ultimately, Hasse diagram is used to express the possible change results and the sensitivity of schemes and indicators is obtained. By changing the weight space, the partial order method can construct a variety of sensitivity analysis methods suitable for global, local and finite. The example application shows that the partial order results of the three kinds of sensitivity analysis are completely consistent with the simulation results, which reflects the effectiveness and uniqueness of the partial order method.

Key words: sensitivity analysis, multiple criteria decision making, weight space, global sensitivity