Int. J. Metaheuristics, Vol. 1, No. 2, 2010 181
Copyright © 2010 Inderscience Enterprises Ltd.
Seasonality and neural networks: a new approach
Bruce Curry* and Peter H. Morgan
Cardiff Business School, Cardiff University,
Aberconway Building, Colum Drive,
Cardiff CF10 3EU, UK
Fax: 44-29-20874419
E-mail: curry@cardiff.ac.uk E-mail: MorganPH@cardiff.ac.uk
*Corresponding author
Abstract: This paper is a response to difficulties reported in applying
feedforward neural networks (NNs) to seasonal data. The solution we propose
is a modified network model which is pruned and optimised by means of
Differential Evolution methods. The problem for NNs in the case of seasonality
lies in the so-called ‘universal approximation’ property, which underpins the
use of MLP networks as a vehicle for flexible non-linear regression. Our view
is that seasonality is best modelled by using sinusoids, which permit the use of
more powerful analytical tools without losing any generality as compared with
dummy variables. However, the actual theorems supporting NN approximation
specifically relate to functions possessing suitable properties of smoothness, in
which case it is not surprising that NNs have difficulty with seasonality. Only a
very ‘short’ sinusoid would be smooth enough. Our suggested solution is to
transform the input variable so that instead of using a time variable alone
we have sinusoids as inputs. In theoretical terms, this helps restore the
approximation property, as can also be seen in our examples, which also serve
to illustrate the strength of Differential Evolution methods.
Keywords: neural network; NN; universal approximation; seasonality; Fourier
series; evolutionary computation.
Reference to this paper should be made as follows: Curry, B. and
Morgan, P.H. (2010) ‘Seasonality and neural networks: a new approach’, Int. J.
Metaheuristics, Vol. 1, No. 2, pp.181–197.
Biographical notes: Bruce Curry is Emeritus Professor of Business Statistics
at Cardiff Business School, Cardiff University, UK. His research interests
incorporate business-related aspects of mathematics, statistics and computing.
He has published many papers in important international journals on topics
such as neural networks and rough sets. His work includes both theoretical
aspects and applications of these topics. He has also acted as a Reviewer for a
wide range of academic journals, across a variety of academic disciplines.
Peter Morgan is currently a Reader in Quantitative Analysis at Cardiff Business
School, Cardiff University, UK. Having been at Cardiff University since 1975
he joined the Business School in 1990 and teaches quantitative methods both to
undergraduates and postgraduates. He holds a BSc (Tech.) in Industrial
Chemistry and PhD in Polymer Physical Chemistry and has a background in
chemical computing. He has also worked in the fields of technology transfer
and innovation skills training. His research interests lie in the fields of neural
networks, evolutionary programming and exploratory data analysis. He is also a
member of the Violence and Society Research Group at Cardiff University.