Title: Physical time series prediction using Recurrent Pi-Sigma Neural Networks

Authors: Abir Jaafar Hussain, Panos Liatsis, Hissam Tawfik, Atulya K. Nagar, Dhiya Al-Jumeily

Addresses: School of Computer and Mathematical Sciences, Liverpool John Moores University, Byrom Street, Liverpool L3 3AF, UK. ' School of Engineering and Mathematical Sciences, City University, Northampton Square, London EC1V 0HB, UK. ' Intelligent and Distributed Systems Laboratory, Deanery of Business and Computer Sciences, Liverpool Hope University, Liverpool L16 9JD, UK. ' Intelligent and Distributed Systems Laboratory, Deanery of Business and Computer Sciences, Liverpool Hope University, Liverpool L16 9JD, UK. ' School of Computer and Mathematical Sciences, Liverpool John Moores University, Byrom Street, Liverpool L3 3AF, UK

Abstract: This paper presents a new type of recurrent neural network, called the Recurrent Pi-Sigma Neural Network (RPSN) and its application to physical time series prediction. The network is constructed of two layers, the sigma and the pi unit layers. The recurrent pi-sigma network calculates the product sum of the weighted inputs and passes the results to a non-linear transfer function. The output of the network is the feedback to its input. The performance of the network is tested in non-linear and non-stationary physical signal prediction. Two popular time series, the mean value of the AE index and the number of sunspots, are used in our studies. The simulation results showed an average improvement in the Signal to Noise Ratio (SNR) of 1.85 dB over the feedforward pi-sigma neural networks.

Keywords: physical time series; forecasting; pi-sigma network; time series prediction; recurrent neural networks; simulation.

DOI: 10.1504/IJAISC.2008.021268

International Journal of Artificial Intelligence and Soft Computing, 2008 Vol.1 No.1, pp.130 - 145

Published online: 14 Nov 2008 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article