Neural, Novel and Hybrid Algorithms for Time Series Prediction

Article excerpt

Timothy Masters, Neural, Novel and Hybrid Algorithms for Time Series Prediction, New York, NY: John Wiley, 1995; Joseph S. Zirilli, Financial Prediction Using Neural Networks, London, UK: International Thompson Computer Press, 1997; and Robert R. Trippi and Jae K. Lee, Artificial Intelligence in Finance & Investing, Chicago, IL: Richard D. Irwin, 1996.

You have heard the claims about forecasting with neural networks, such as: "Any problem that can be solved with traditional modeling or statistical methods can most likely be solved more effectively with a neural network." If you are still making excuses about why you have not tried forecasting with neural networks, three recent books will blow away your excuses.

Masters included NPREDICT, a time series forecasting program (in DOS and Windows NT versions), on the CD-ROM included with his book. Further, he supplied the C++ source code so that you may modify his system. From easy moving average methods to Box-Jenkins methods, Masters demonstrated the superiority of neural network and of hybrid neural network techniques. For validation, he used practically distribution-free confidence intervals.

Forecasters will agree with the reasons Masters cited for writing his book: (a) Many neural networks books are theoretical only. (b) Other neural network books are broad overviews and lack implementation details. His book is required reading for professional forecasters.

Joseph Zirilli, owner of MJ Futures and distributor of BrainTrader System Builder and of NeuroGen (a Windowsbased neural network application), has tested his systems in the day-trading jungle of institutional traders (futures, commodities, stocks, bonds and currencies). In the pits, forecasting validation consists of walk-forward testing and rprofit%. The trader's approach includes consideration of initial account size, commissions, and of drawdowns. While Zirilli's 135-page book focused on futures markets, his methods remain appropriate for corporate financial forecasting. …