作者: George Tselentis , George Dounias
DOI: 10.1007/978-1-4757-2845-3_7
关键词:
摘要: A Neuro-fuzzy windowing scheme is applied to predict exchange rates. By using a hybrid learning procedure an input-output mapping can be constructed based on training data pairs. The applies the ANFIS (Adaptive-Network-based Fuzzy Inference System) algorithm which least-squares method and back-propagation gradient descent for identifying linear nonlinear parameters, respectively, in Sugeno-type fuzzy inference system. retraining performed moving window of input data, thus keeping track latest changes. An evaluation short-term prediction interval real rates USD vs. GRD, DEM, FRF ECU.