Optimization of the Interval Type-2 Fuzzy Integrators in Ensembles of ANFIS Models for Time Series Prediction: Case of the Mexican Stock Exchange

Abstract

This paper describes the optimization of the fuzzy integrators in Ensembles of ANFIS model for time series prediction: case of the Mexican Stock Exchange. The Mexican stock exchange that is used corresponds to the period of 11/09/2005 to 01/15/2009 to simulate the performance of the proposed architecture. We used interval type-2 fuzzy systems to integrate the outputs (forecast) of each of the ANFIS models in the Ensemble. Genetic Algorithms (GAs) are used for the optimization of memberships function "MFs" for the 2 MFs (used linguistic labels "Small and Large") and for the 3 MFs (used linguistic labels "Small, Middle and Large") parameters of the fuzzy integrators. In the experiments the genetic algorithms optimized the Gaussian, Generalized Bell and Triangular membership functions for each of the fuzzy integrators; in the interval type-2 fuzzy integrator there are more parameters, thereby increasing the complexity of the training for the fuzzy integrators. Simulation results show the effectiveness of the proposed approach in comparison with other researchers.