This paper focuses on the problem of supplier portfolio selection where a company has to choose the best possible set of suppliers with respect to various constraints. An intuitive heuristic can suggest to use any of the methods for suppliers ranking and then to put the first one into the portfolio. If some required constraint is not met, then the second supplier according to the ranking is added, and so on, until all the constraints are satisfied. However, such approach can result in a non-optimal decision. The constraints can cause that a combination of the alternatives with lower rankings can be better, than some higher-ranked alternative from the perspective of feasibility. To build the optimization model, the authors of this paper use the PROMETHEE V method: a popular combination of multi-criteria decision making method PROMETHEE and mixed integer programming. However, it is shown that the original PROMETHEE V method, namely the logic under which an objective function is set, is not suitable here and leads to discrimination of suppliers with worse ranking. Therefore, a modification, which brings more reasonable results, is proposed in this paper. A numerical example is used to show the suitability of the proposed approach and compare the results with the original algorithm and also with one prior modification introduced by by other authors in the past. The analysis is further supported by a thorough sensitivity analysis using flexible and parametric programming.
Read full abstract