Authors: Gary R. Weckman, Sriram Lakshminarayanan, Jon H. Marvel, Andy Snow
Addresses: Industrial and Systems Engineering, Ohio University, Stocker Center 280, Athens, Ohio 45701, USA. ' Mcube Investment Technologies, 5240 Tennyson Parkway, #102, Plano, Texas, 75024, USA. ' Department of Management, Gettysburg College, 300 North Washington Street, Box 395, Gettysburg, Pennsylvania, 17325, USA. ' The J. Warren McClure School of Information and Telecommunication Systems, Ohio University, Lindley Hall 291, Athens, Ohio 45701, USA
Abstract: This paper focuses on the development of a stock market forecasting model based on artificial neural network architecture. A baseline neural network model was developed using GFF architecture. The performance of the baseline model was evaluated by using representative large-cap stocks in six critical industrial sectors. Key performance measures, which included correlation coefficient and mean square error, were identified and used to compare the different models. A self-organising map network was developed to reduce the set of 56 stock market indicators into a final set of 11 indicators that covered market momentum, market volatility, market trend, broad market indictors and general momentum indicators. The model still required additional developments to better forecast turning points in the market. Based on Elliot|s Wave Theory, two additional indicators were introduced to improve the forecast accuracy for turning points.
Keywords: artificial neural networks; ANNs; stock market forecasting; self organising maps; fuzzy set theory; technical indicators; generalised feed forward networks; turning points; stock markets; performance measures; market momentum; market volatility; market trends; broad market indictors; general momentum indicators; forecast accuracy.
International Journal of Business Forecasting and Marketing Intelligence, 2008 Vol.1 No.1, pp.30 - 49
Published online: 17 Oct 2008 *Full-text access for editors Access for subscribers Purchase this article Comment on this article