[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Code for Pattern Recognition: Head and Shoulders



PureBytes Links

Trading Reference Links

A bit belated on this thread, but I thought I would give those looking
for something interesting on a Sunday a possibility. Here is a link to
some code for pattern recognition of a head and shoulders top. 

http://www.tarnsoft.com/headandshoulder.html

Not the be-all and end-all in pattern detection. But a valid
starting point for developing a pattern detection algorithm. Certainly
picked up the head and shoulders top in IBM very nicely. 

I'm not going to repeat all that was in my prior plugs, but here's the
website again (free 30 day no-strings no-personal-information trial --
complete system editing-testing capability). 

http://www.tarnsoft.com/

-- Tom Kohl (Tarn Software)

PS we do want to hear from potential realtime beta testers. 


Bill Vedder wrote:
> 
>         There has been some interest in pattern recognition on the list
> lately.  In that vein, some folks at the NY Fed have published quite a
> bit of work looking at the "head and shoulders" pattern.  Carol Osler
> has been doing this since 1994, at least. If interested (the few, the
> proud, the brave...) follow this link:
> 
> http://www.frbsf.org/system/fedinprint/index.html
> 
> Enter "osler" in the AUTHOR section of the search input (or try
> "shoulder" in the KEYWORD section). It should take you to a list of
> about 4 or 5 of her papers. I know the 1998 paper is available online in
> pdf format. The others (and they appear more interesting) will probably
> require a trip to the library.
> 
> It might be interesting to see, at least, how they go about defining the
> pattern. They may even  show some computer code, eh?
> 
> Best Regards,
> 
> Bill Vedder
-----------------------
"M. Simms" wrote:
> 
> Creating computer code to do this is non-trivial....it's like creating a
> special form of OCR software....for reading lines instead of black dots.
> It took about 4 years of incredibly complex code to finally make OCR
> reliable enough to be useful.
>