Fuzzy modeling of nonlinear stochastic systems by learning from examples
Amir H. Meghdadi, Mohammad-R. Akbarzadeh-T · 2002
The conventional fuzzy logic techniques have been extensively used in modeling nonlinear and complex systems. Such techniques, however, generally ignore the statistical nature that many complex systems may exhibit. When the system's behavior is significantly influenced by stochastic parameters, it is reasonable to expect that the modeling performance would be improved if the effect of such parameters is taken into account. In this paper, a novel modification to a table look-up scheme is proposed. A new stochastic chaotic time series is also introduced. It is demonstrated that the modified method improves learning accuracy by considering the system's stochastic nature. Moreover, examining the results at different noise levels reveals that this improvement is indeed due to the stochastic nature of the system.