An evolving type-2 neural fuzzy inference system

Traditional designs of neural fuzzy systems are largely userdependent whereby the knowledge to form the computational structures of the systems is provided by the user. By designing a neural fuzzy system based on experts' knowledge results in a non-varying structure of the system. To overcome the drawback of a heavily user-dependent system, self-organizing methods that are able to directly utilize knowledge from the numerical training data have been incorporated into the neural fuzzy systems to design the systems. Nevertheless, this data-driven approach is insufficient in meeting the challenges of real-life application problems with time-varying dynamics. Hence, this paper is a novel attempt in addressing the issues involved in the design for an evolving Type-2 Mamdani-type neural fuzzy system by proposing the evolving Type-2 neural fuzzy inference system (eT2FIS) - an online system that is able to fulfill the requirements of evolving structures and updating parameters to model the non-stationeries in real-life applications.