Time Series Forecasting by Evolving Artificial Neural Networks Using "Shuffle", Cross-Validation and Ensembles

Accurate time series forecasting are important for several business, research, and application of engineering systems. Evolutionary Neural Networks are particularly appealing because of their ability to design, in an automatic way, a model (an Artificial Neural Network) for an unspecified nonlinear relationship for time series values. This paper evaluates two methods to obtain the pattern sets that will be used by the artificial neural network in the evolutionary process, one called "shuffle" and another one carried out with cross-validation and ensembles. A study using these two methods will be shown with the aim to evaluate the effect of both methods in the accurateness of the final forecasting.

• Type: Journal
• Year: 2010
• Where: ICANN
• Authors: Juan Peralta, Germán Gutiérrez, Araceli Sanchís
• Keywords: Artificial Neural Network, Evolutionary Neural Networks, ICANN 2010, Neural Network, Neural Networks