Evolutionary design of efficient type-2 fuzzy models from noisy data using hybrid PSO model

This work presents an efficient framework for designing type-2 fuzzy models from noisy data that represent majority of the real-world databases used for building models. The framework is based on the use of a hybrid particle swarm optimisation (PSO) model that combines the traits of gbest and lbest PSO variants for getting fast results and at the same time avoiding premature convergence. Two major components of the framework, i.e., the encoding mechanism, and the hybrid PSO model have been discussed. Further, two benchmark problems that cover major application domains of type-2 FLSs viz. forecasting and classification, have been presented to demonstrate the effectiveness of the proposed methodology. For each noisy benchmark data, type-1 fuzzy model is also designed and its performance is compared with the designed type-2 fuzzy model with an objective to exhibit better noise handling capability of type-2 models.