An Analysis of the Application of B-spline Neurofuzzy Construction Algorithms
Monika Rose Brown, K.M. Bossley, Chris J. Harris · ePrints Soton (University of Southampton) · 1996
This paper investigates the application of the B-spline neurofuzzy construction algorithms, developed by the ISIS research group, to identifying appropriate models from two standard data sets. The two data sets are the Box-Jenkins data and a simulated system data set, both of which can are described by discrete time series models. It is shown that the construction algorithms pick out the dominant linear components associated with this data, as well as the small nonlinearities, and the model's structure facilitates the analysis of the data where errors occur. This is simplified by the incorporation of these techniques inside NeuFrame: a PC-based graphical development environment.