|
Resumen:
|
A fully precise numerical evaluation of disasters’ effects is unrealistic in the time-pressured, highly uncertain decision context taking place just after a disaster strike. This is mainly due to some features of the available information in such a context, but also because of the imprecise nature of some of the relevant categories (think about the number of affected people, for instance). Instead of a numerical evaluation, in this work is considered that it is rather more plausible and realistic to classify the severity of the consequences of a disaster in terms of the relevant scenarios for the NGO’s decision makers. Therefore, the abovementioned practical problem of evaluation of disaster consequences leads to a classification problem in which the classes are identified with the linguistic terms that describe those relevant scenarios. In order to carry out this classification and ensure the linguistic adaptation and the understandability of the proposed solution, the methodology of the descriptive fuzzy rulebased classification systems has been adopted in this work. Nevertheless, some features of that context, as the ordering and gradation of the consequences or the need of avoiding the risk of underestimation of the effects of disasters, entail the necessity of considering and assuming an structure over the set of classes or linguistic labels, somehow modeling those features inside of the classification model. Such an structure is introduced here by means of the notion of dissimilarity between classes, leading to a bipolar knowledge representation framework which allows to adequate the classification models to the constraints and requirements of the NGO context.
|
