|
PureBytes Links
Trading Reference Links
|
David,
I've heard that GAs have had some success, but I have no
first hand experience. It's on my "list" of things to
investigate.
You do make an important distinction about making "one's targets
clear" and give the example of predicting turning points. I
do know of some success with predicting turning points with NNs.
I should have made the distinction that I don't know of any
NN-based models that have done much better at price forecasting
than you could obtain by a random walk model. I have
a bit of firsthand knowledge of this, I worked for a quant
firm at the CBOT in the late 80s... this is at a time when
NNs were all the rage... and continued that way into the 90s.
Folks were treating NNs as blackboxes that might instantly
give them an edge. The rage gradually faded...
My intent here is not to be overly-pessimistic or confrontational,
but rather I want to point out that with NNs in particular
(and I presume also with GAs) there is a lot of preprocessing
of data and understanding of the underlying phenomenon necessary
in order to get decent results.
I have tried to apply to the markets my own NN [which
was originally developed for geophysical applications:
http://webserv.gsfc.nasa.gov/ESD/ESU/1997.01.24.ESU.html#1]
but I so far have had better success in the markets using
"simpler" models.
Jeff
David Jennings wrote:
>There are quite a few instances of success trading with either GAs or NNs. I
>must confess, I found The Predictors a bit populist. I'd hoped to get some
>insighst into model structure. Quite a number of the banks put up big money
>for research into fuzzy networks, GARCH and ARCH. $5M p.a. was a typical
>research budget.
>
>The ultimate Trading Guide is worth a read on how good GAs and NNs are. I
>must admit I'm a great fan of GAs for seeking optimal parameters for an
>indicator over a portfolio. Model structure is very important for trading a
>market - whatever chaos theory tells us, and thus one can end up with an
>overly complex model which misses the point.
>
>My experience of NNs is that one has to be quite specific about one's target
>e.g. predicting turning points etc.and then use them in combination. I'm
>convinced that a guy with sound experience in say Fib could do as good a job
>as any NN. however, Lionel makes a strong point though, the beauty of AI in
>general is that it can play multiple markets in multiple time frames
>simultaneously - and hence generate super normal gains or losses. There was
>a group of guys from Rice university (whose names I've forgotten) who
>launched Frontier specialising in pattern recognition. They built trading
>systems and played the markets. The very absence of research papers from
>them suggests that they were successful!
>
>DJ
>> ----- Original Message -----
>> From: "Jeff Haferman" <haferman@xxxxxxxxxxxxxxxxxxxxxxx>
>> To: <metastock@xxxxxxxxxxxxx>
>> Sent: Tuesday, February 06, 2001 11:51 AM
>> Subject: Re: Artificial Intelligence, expert and neural
>>
>> >
>> > But you're going to spend a lot of time training and tuning any
>> > expert system or neural net. The results I've seen indicate that
>> > neural nets are great at learning about past market activity,
>> > but I haven't seen any convincing results that show that
>> > they do better than random at predicting future activity.
>> >
|