Call for Papers
Special Issue of the International Journal of Forecasting on
“Forecasting with artificial neural networks and computational intelligence”
Motivation & Context
The last 20 years of research have produced more than 5000 publications on artificial neural networks (NN) for predictive modelling across various disciplines. However, while NN and other methods of computational intelligence (CI) are firmly established in automatic control and classification problems, they have not received the same level of attention in time series forecasting (regression). Many of the optimistic publications indicating a competitive or even superior performance of NNs have focussed on theoretical development of novel paradigms, or extensions to existing methods, architectures, and training algorithms, but have lacked a valid and reliable evaluation of the empirical evidence of their performance. Similarly, only a few publications have attempted to develop a thorough methodology on how to model NNs under specific conditions, limiting the modelling process of NNs to a heuristic and ad-hoc ‘art’ of hand-tuning individual models, rather than a scientific approach using a replicable methodology and modelling process. As a consequence, NNs have not yet been empirically validated as a forecasting method in many areas of forecasting, despite theoretical advances.
To explore this gap between academic attention, theoretical prowess and empirical performance we invite contributions to a special issue of the International Journal of Forecasting (IJF) dedicated to evaluating the evidence on forecasting with NN and CI-methods.
Papers for this special issue should focus on novel techniques, methods, methodologies and applications from the computational intelligence domain, with particular emphasis on neural networks, within all aspects of forecasting. Particular emphasis will be placed on applied or applicable work that provides valid and reliable evidence on the performance of the methods and the development of robust methodologies based upon rigorous evaluation, rather than purely theoretical contributions. Contributions of contenders that have contributed to one of the recent forecasting competitions dedicated to NN and CI-methods (ESTSP’07, ESTSP’08, NN3 and NN5) are particularly encouraged. Due to the single-time origin design of these competitions, the authors are encouraged to obtain the complete datasets and rerun experiments for their papers, in order to obtain representative out-of-sample results across multiple origins and error measures in comparison to established statistical benchmark methods, adhering to the best-practices set out in discussions in the IJF (see e.g. Tashman (2000) Out-of-sample tests of forecasting accuracy - an analysis and review, International Journal of Forecasting 16, 437–450; and Adya and Collopy (1998) How effective are neural networks at forecasting and prediction? A review and evaluation, Journal of Forecasting, 17, 481–495).
About the Journal
The International Journal of Forecasting (IJF, www.forecasters.org/ijf) published by Elsevier is the leading journal in its field and indexed by all major citation indexing services (including Thomson Scientific). It is the official publication of the International Institute of Forecasters (IIF) and shares its aims and scope. The IJF publishes high quality refereed papers covering all aspects of forecasting. Its objective (and that of the IIF) is to unify the field, and to bridge the gap between theory and practice. The intention is to make forecasting useful and relevant for decision and policy makers who need forecasts. The journal places particular emphasis on empirical studies, evaluation activities, implementation research and ways of improving the practice of forecasting. It is open to many points of view and encourages debate to find solutions for problems facing the field. Regular features of the IJF include research papers, research notes, discussion articles, book reviews, and software reviews. The IJF has an impact factor of 1.409 (Journal Citation Reports® 2008, published by Thomson Scientific).
Each submitted paper will be peer-reviewed in the same manner as other submissions to the IJF. Providing papers fit into the theme of the special issue, quality and originality of the contribution will be the major criteria for each submission. Due to the tight deadlines, any paper for which the outcome of the refereeing process is “major revision” will not be included in the special issue, but may be revised and resubmitted according to the journal’s regular process. It may also be considered for a forthcoming special volume on Advances in Forecasting with Computational Intelligence by Springer (which is circulated separately).
Deadline for manuscripts: 15 September, 2008
Preliminary decision to authors: 24 November, 2008
Revision Due: 12 January, 2008
Final Manuscript Due: 16 March, 2008
Authors are encouraged to contact one of the editors with an extended abstract of three pages to discuss any questions of suitability. Only email submissions will be accepted. Please submit your manuscript to IJF_Special_Issue@xxxxxxxxxxxxxxxxxxxxxx. The submission must be in PDF format. Final manuscripts must be submitted in either MS-Word or LaTeX format for typesetting by the publisher. Manuscripts must be in English and double-spaced throughout. Papers should in general not exceed 6,000 words. All submissions will be peer reviewed. Detailed instructions for authors are at: http://www.forecasters.org/ijf.
Prof. Fred Collopy
Information Systems Department
Weatherhead School of Management
Case Western Reserve University
Cleveland, Ohio 44106-7235
Dr. Sven F. Crone
Lancaster University Management School
Research Centre for Forecasting
Lancaster, LA1 4YX
Dr. Amaury Lendasse
Helsinki University of Technology
For general enquiries please contact us via email at: IJF_Special_Issue@xxxxxxxxxxxxxxxxxxxxxx