Comparison of Genetic Algorithm and Particle Swarm Optimization of Ensemble Neural Networks for Complex Time Series Prediction

Abstract

In this paper the development of a method for the design of the general architecture of ensemble neural networks for time series prediction is presented. This method consists on the optimization of the ensemble neural network with response integration based on type-1 and type-2 fuzzy logic. The optimization algorithms that are used to illustrate the proposed method are the genetic algorithm and the particle swarm optimization algorithm and the purpose of applying these methods is to find the best possible architecture of the neural network ensemble with the lowest prediction error. Simulation results with several complex times series are presented to show the advantage of the proposed method, also comparisons are made with other time series, such as Mexican Stock Exchange, and Dow Jones Stock Exchange.