Time series prediction by artificial neural networks and differential evolution in distributed environment

Abstract

Current work will present a model for time series prediction by the usage of Artificial Neural Networks (ANN) trained with Differential Evolution (DE) in distributed computational environment. Time series prediction is a complex work and demand development of more effective and faster algorithms. ANN is used as a base and it is trained with historical data. One of the main problems is how to select accurate ANN training algorithm. There are two general possibilities - exact numeric optimization methods and heuristic optimization methods. When the right heuristic is applied the training can be done in distributed computational environment. In this case there is much faster and realistic output, which helps to achieve better prediction. © 2012 Springer-Verlag.