F Baghbani - Department of Electrical and Computer Engineering, Semnan University, Semnan, Iran
Mehraneh Hemmati - Department of Electrical and Computer Engineering, Semnan University, Semnan, Iran

Type-۲ fuzzy systems have three-dimensional membership functions and a footprint of uncertainties. Hence, they could handle higher levels of uncertainties in comparison with type-۱ fuzzy systems. Here, we employ interval type-۲ Takagi-Sugeno-Kang (TSK) fuzzy systems for system identification and time-series prediction. The designed type-۲ fuzzy system uses direct defuzzification which avoids the computationally extensive calculations of the Karnik-Mendel algorithm. A type-۱ fuzzy system is also designed in the same way for comparative purposes. The weights of the fuzzy systems are updated using the gradient-descent algorithm. The performance of the two fuzzy systems is evaluated in chaotic time-series prediction. Simulation results show that the type-۲ fuzzy system reaches lower error for different levels of measurement noise

Interval type-۲ fuzzy systems, direct defuzzification, gradient descent algorithm, Takagi-Sugeno-Kang fuzzy systems, time series prediction .