|
Résumé :
|
Multiple response problems encompass three stages: data gathering, model building and optimization. Most of existing approaches in literature don't incorporate the obtained effects of modeling stage while these results can be employed in optimization. Moreover, these approaches subjectively increase/decrease the responses to obtain more solution in interactive methods. Hence, this paper proposes least significant difference (LSD) criterion as agent of modeling stage to trade off between responses, which guides the decision makers (DMs) how to change membership function, and the criteria of modeling fitness, i.e. R2, PRESS and Adequate prediction, which guide DMs to order the changes of membership function in an interactive manner. At each iteration, the proposed method changes the membership function of each response by LSD, and solves the modified model. In practice, a decision has to be taken by the members of the management in such a manner that each individual is satisfied with the decision when a company wants to take a decision of specify quality characteristics of products. Thus, views of points of DMs about the resulted solutions of iterations are taken, and then a fuzzy group decision making approach is applied to select the preferred solution among the resulting solutions. Finally, we demonstrate our method by a numerical example and show the proposed approach is not an invalid method in comparison with others in literature.
|
