Multiple criteria optimization problems deal with a type of decision making, where the final choice is affected by multiple criteria. This thesis is devoted to a variety of methods, with which such problems can be solved. A typical feature of all the considered methods is that the multiple criteria problems are converted into single criterion ones. As a basis we take the simple additive weighting method, where the weights are known already at the start. The analytic hierarchy process takes into account also the decision-maker's preferences, which directly influence the final choice. The simplest case treats a special type of positive reciprocal matrices, which are called consistent. Perron-Frobenius theorem tells us that these matrices have a positive eigenvector, which after normalization is taken as the weight vector. In more complex cases we define the random index and the consistency index, which are then used to define the consistency ratio of the matrix. If the matrix is not consistent enough, we modify it appropriately. Finally we consider fuzzy analytic hierarchy process, where the decision-maker's preferences are not certain and are modeled using fuzzy numbers. Each of the methods is applied to solve a practical example, for which all the methods produce comparable results.
|
