|
PureBytes Links
Trading Reference Links
|
On Fri, 14 Jul 2000 17:20:39 EDT, JMMain1000@xxxxxxx wrote:
>Does anyone on the list have any public code that will take a fast fourir
>transform of data and also the inverse fast fourir transform - in any
>coding language. EXCEL can do it but I want something that can
>be ported to a DLL.
>
>I read with interest Dennis Meyers article in Futures Magazine about trading
>the intraday S and P futures with the Fast Fourir transform and dsing walk
>forward analysis.
>
>I someone would like to join in such a project, please contact me.
>
>Jim
Jim,
There is in Chapter 11 of "The Profit Magic of Stock Transaction
Timing" by J. M. Hurst, a description of a method of Fourier that I
have thought could be coded into an ELA. There is a (available on AOL
do a search for FFT or Fourier) shareware program available, which I
can send you if you want. There are a couple of others, Check The
Foundation for the Study of Cycles, they have one for $450 and another
for around $900. There is a DOS based one from Dynacomp Inc check
www.dynacompsoftware.com which I own, costs about $170 and is designed
for stocks and commodities. It will make a price projection based on
FFT from an ASCII data file you have input, and has an educational
manual of about 100 pages. You will need to give a lot of
consideration to how you will taper your data so that the FFT will not
try and give you the frequency for the spikes that occur at the start
and end of your data (the OEX for example will jump from 0 to say 250
and then this week from 800 to 0). The FUTURES articles give you a
start on this. Having given you all the above I would strongly advise
you to get and read Hurst before you go any further. Once you get an
FFT or DFT to do as much for you as they can you will find that it is
just a kind of "suggestion" about phase, period and amplitude of the
ideal. Hurst will help you with the practical application. You will
find that a human with only a minor amount of pattern recognition will
give you a better cycle analysis (for market prices anyway) than a
very advanced FFT program.
I would of course be interested in communicating with you, and others
interested in this area of research.
But, you asked for code, not canned programs. The following url:
http://www.fftw.org/links.html
Will yield the following info (but with the links active):
FFT Source Code
The following are places where you can download source code for FFTs.
(There are so many FFT implementations available that we mostly link
to sites that are themselves collections of code or links.)
The FFTW Home Page:(www.fftw.org/) A fast C library for performing the
FFT in one or more dimensions, including parallel and real-complex
transforms. Of course, we have to include ourselves in this list!
FFT Sources: This is the list of all the codes that we included in
benchFFT, along with links to where they may be downloaded. It is one
of the more complete FFT-software listings available.
Jörg's "Ugly" Page: Jörg Arndt has gathered a menagerie of FFT links
and source code, including much of the software that we used in our
benchmark. (Somewhat chaotically organized.)
An FFT Page maintained by Steve Kifowit, focusing primarily on Fortran
code.
D. J. Bernstein has written an optimized C FFT for the Pentium and
UltraSPARC. (Its output is permuted and it is limited to N <= 8192,
but this doesn't prevent it from being used e.g. for convolutions. It
includes convolution routines for real and complex data.) (See also
pfftw.)
PD FFT Page: A collection of several public-domain FFT codes and other
links.
"Success is the only test of genius" -R.Adm Daniel Gallery 1901-1977
|