摘要：AbstractIn the fuzzy neural networks’ applications the membership functions μ_A (t) are selected from a standardized sequence of fuzzy sets without any modification depending on the class of problems to be solved using these procedures. In this paper we are creating fuzzy numbers strictly depending on the studied stock's problem. The dependence on the problem is realized determining the member – ship function μ_A (t), using the least squares method containing the relative frequencies of the stock's outputs. Is well known that the least squares method is a best approximation procedure. Therefore the whole numerical procedure applied in the followings using the (FAHP) and (FQFD) algorithms depends on the solved problem's data. Moreover (FAHP) proposes the best weights we’ll use in (FQFD) weights depending on the input data.
