| || || Nand, Ravneil|
| || || Competitive Island cooperative neuro-evolution for time series prediction |
Institution: University of the South Pacific.
Subject: Evolutionary programming (Computer science), Evolutionary computation
Call No.: pac QA 76 .618 .N36 2016
Copyright:Over 80% of this thesis may be copied without the authors written permission
Abstract: Cooperative coevolution is an evolutionary computation method used for Neuro-evolution. It breaks down neural network topology into smaller parts using problem decomposition and evolves their weights. Competition and collaboration can take advantage of the strengths of the different methods and eliminate their limitation. Feature selection of time series problems produces reconstructed state-space vector that is used for training neural networks. In this thesis, a competitive cooperative coevolution method is applied for the Neuro-evolution of feedforward networks for time series prediction. Furthermore, the competition based approach uses two different data sets generated from same reconstructed state-space vectors acquired from the same time series data. The results show that the proposed approaches have improved the results when compared to related methods in the literature.