Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Neural Nets - A tool for the 90's -- Ignore unavailable to you. Want to Upgrade?

To: Tim Krieg who wrote (657)	9/10/1999 9:37:00 AM
From: Optim	Read Replies (2) \| Respond to of 871

Tim, Thanks for the message. Sorry it has taken me so long to acknowledge it, but to be honest I didn't understand what you talking about until this week... :o) I have spent the last two weeks or so experimenting with Wavelets and reading up on FFT and similar methods. I bought a book about a week ago, The World According to Wavelets : The Story of a Mathematical Technique in the Making from Barbara Burke Hubbard (ISBN: 1568810725 for anyone interested). It does a great job of explaining FFT and Wavelets without getting too technical. The book is split in two and the first part explains the theory and applications in simple terms. The second part contains more of the mathematical content to explain the concepts further for those that wish to read it. I must admit that my understanding is very basic so far, so I am not sure I am applying these tools in the proper fashion. However I have come up with some great ($$$!) models using only 3 or 4 inputs based on Morlet Wavelets of the OHLC. An example S&P model made 190% annualized return on margin out-of-sample for 1 1/2 years, trained on 3 years of data. Another model for CSCO made 240% annualized for 9 months out of sample, and has held up well for the last week or so. I still am unsure if I should complement the wavelet inputs with other indicators to explain trend, support and resistance, etc. From my understanding so far FFT and wavelets reduce a series to its basic underlying cycles, isolating them by frequency. It seems the neural network loves these inputs as there is much less noise and it can do a much better job predicting a few bars out. Just out of curiosity, what software do you use for your FFT work? I think Excel has a FFT function but I haven't used it so far. I have looked at the FFT functions available in NS Trader but I am unsure as to how to apply them properly. I have much more experimenting and research to do. For anyone else that is interested, Technical Analysis of Stocks and Commodities (TASC) ran a few articles on Fourier Transforms in April and May of 1999. They give a pretty decent basic explanation. Thanks for the help! Optim