A self-learning fuzzy modeling approach with its application to EEG time-series prediction problem

A self-learning fuzzy modeling approach based on TSK model is proposed in this paper. Based on the input-output training data, the fuzzy system optimizes the linear parameters in the THEN part of the fuzzy rules using the steady-state Kalman filter and the membership function parameters in the IF part by using supervised Gaussian learning rule. The application to EEG time-series prediction has demonstrated the practical effectiveness of the approach proposed.