Use of artificial neural networks in initial ponderation of AHP techniques applied to analysis of watershed vulnerability

In many decision problems, the information provided by decision makers is often inaccurate or uncertain due to the time constrains, the lack of data or capacity to dealing with available information. In order to overcome these difficulties, some studies have shown the application of the technique Analytical Hierarchic Ponderation (AHP) in the geosciences. This technique breaks a complicated decision problem into a hierarchy of sub-problems, makes comparisons pairwise of the input information and from the results, allow estimating weights for each considered variable. However, this study recognize that the weakest point of the method is the initial weighting of the input variables, as the expert knowledge can be subjective and relative importance of the elements of analysis may vary. In the literature, there is no discussion about the effects of these weights and their variation on final weights. Therefore, this research evaluates different scenarios, considering the environmental context of vulnerability, comparing the adjustment of initial ponderation values of common AHP in relation of expert's ponderation and in relation to initial weights obtained from neural network in order to minimize the subjectivity of the analysis process

Keywords:
Artificial Neural Networks; Fuzzy AHP; Environmental Vulnerability