Evolutionary computation based identification of a monotonic Takagi-Sugeno-Kang fuzzy system

This paper introduces an evolutionary computation (EC)-based identification method of a Takagi-Sugeno-Kang (TSK) fuzzy system constrained by monotonic input-output relationship. The differentiation of a TSK fuzzy system output with respect to its input yields a sufficient condition of the fuzzy system parameters that makes the fuzzy system monotonic. By using the derived condition, we suggest a new EC-based fuzzy system identification method whose fuzzy model preserves monotonicity at every identification stage by means of modified representation and mutation paradigms. Simulation results show that the proposed identification technique is better than conventional methods in its convergence rate, generalization characteristic, and robustness.