<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Hi Pierre<br>
<br>
Don't want to sound unhelpful but I'm afraid my employer would't
want me to go further. <br>
<br>
Have a look at Chapter 5 of my PhD Thesis
<a class="moz-txt-link-freetext" href="http://www.xesoftware.com.au/ThesisAsPassed.pdf">http://www.xesoftware.com.au/ThesisAsPassed.pdf</a><br>
<br>
It is entitled <i>Experiments in Classification</i> and takes you
through R and the various techniques. True its in a rather
different domain (speech recognition errors) but the problem is just
the same: one of classification.<br>
<br>
<div class="moz-signature">Stephen Choularton Ph.D., FIoD<br>
<br>
9999 2226<br>
0413 545 182<br>
<br>
</div>
<br>
On 24/08/2011 7:00 PM, Gentil Homme wrote:
<blockquote
cite="mid:CAM9XErB3JXSK4TTCQpDP+iQy3jEWXVJfLawbyJQ1ezT_Q3zC=Q@mail.gmail.com"
type="cite">Thanks Stephen,<br>
<br>
<br>
I'm currently first attempting to define one architecture (inputs
to NN, which I believe need some preprocessing/denoising, ....)<br>
I think that the process to train and test NN performance must be
well organised, and the validation steps are much critical ...<br>
<br>
I found also the neuralnet package that seems interesting for NN
training.<br>
<br>
I'll look after svms in a second step.<br>
<br>
Which kind of data did you use in your nn tests ?<br>
<br>
<br>
Best Rgds,<br>
<br>
Pierre<br>
<br>
<br>
<br>
<div class="gmail_quote">2011/8/24 Stephen Choularton <span
dir="ltr"><<a moz-do-not-send="true"
href="mailto:stephen@organicfoodmarkets.com.au">stephen@organicfoodmarkets.com.au</a>></span><br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div bgcolor="#FFFFFF" text="#000000"> I use nnet and it
appears to work reasonably 'out of the box' as an algorithm
but you get better performance if you play with the
parameters to tune it.<br>
<br>
svm appears to work best when you select Gaussian so I'm not
sure about your assumption.<br>
<br>
All these algorithms have similar R calls so its not to
difficult to try lots of them out. Try then out of the box
first up.<br>
<br>
good luck.<br>
<br>
<div>
<div class="im">Stephen Choularton Ph.D., FIoD<br>
<br>
</div>
9999 2226<br>
0413 545 182<br>
<br>
</div>
<div>
<div class="h5"> <br>
On 23/08/2011 4:51 PM, Gentil Homme wrote: </div>
</div>
<blockquote type="cite">
<div>
<div class="h5">I didn't know about the theorem but it
seems reasonable to believe that some techniques are
more appropriate than others for modelling/predicting
financial data.<br>
It should be because of their nature : non gaussian,
non linear, non stationary, ...<br>
<br>
I think it's like the usual technical indicators
(MACD, Stochastic, etc ... ) which are more or less
suitable depending on the market conditions.<br>
<br>
What would be your recommended R package for NNs, as
there are different possible architecture : GRNN, PNN,
SOM ... (see <a moz-do-not-send="true"
href="http://en.wikipedia.org/wiki/NeuroSolutions"
target="_blank">http://en.wikipedia.org/wiki/NeuroSolutions</a>)<br>
<br>
Before trying many solutions, maybe it's worth to have
some discussion ... there can be another mighty
theorem we should all know :-)<br>
<br>
Best Rgds,<br>
<br>
Pierre<br>
<br>
<br>
<br>
<div class="gmail_quote">2011/8/23 Stephen Choularton
<span dir="ltr"><<a moz-do-not-send="true"
href="mailto:stephen@organicfoodmarkets.com.au"
target="_blank">stephen@organicfoodmarkets.com.au</a>></span><br>
<blockquote class="gmail_quote" style="margin:0pt
0pt 0pt 0.8ex;border-left:1px solid rgb(204, 204,
204);padding-left:1ex">I think that Gentil should
be aware of the /No Free Lunch Theorem/ (Duda et
al., 2001, Wolpert and Macready, 1997). There are
no context-independent or usage-independent
reasons to favor one machine learning algorithm
over another. If one performs better than another,
it is owing to its better fit to the particular
problem, not its general superiority. If you wish
to use these techniques try lots of them:
certainly neural networks and support vector
machines, but also try some of the ensemble
techniques such as bagging, boosting and random
forest. You can even try the statisticians
favorite, logistic regression. They are all
available in R.<br>
<font color="#888888"> <br>
Stephen Choularton Ph.D., FIoD</font>
<div>
<div><br>
<br>
<br>
On 23/08/2011 12:14 AM, Brian G. Peterson
wrote:<br>
<blockquote class="gmail_quote"
style="margin:0pt 0pt 0pt
0.8ex;border-left:1px solid rgb(204, 204,
204);padding-left:1ex"> On Mon, 2011-08-22
at 10:11 +0200, Gentil Homme wrote:<br>
<blockquote class="gmail_quote"
style="margin:0pt 0pt 0pt
0.8ex;border-left:1px solid rgb(204, 204,
204);padding-left:1ex"> I just send out
this post in order to share within
r-sig-finance any<br>
possible experience, advice, ... about NNs
or SVMs with R.<br>
</blockquote>
It seems that you're asking us to share with
you, and not sharing much<br>
yourself in return.<br>
<br>
Perhaps you could answer your own questions
in this thread with the<br>
things you are trying?<br>
<br>
SVM's have been discussed on this list many
times, please search the<br>
list archives.<br>
<br>
This blog has covered this topic:<br>
<a moz-do-not-send="true"
href="http://www.aphysicistinwallstreet.com/"
target="_blank">http://www.aphysicistinwallstreet.com/</a><br>
<br>
Also, there are a few books on machine
learning that use R.<br>
<br>
<blockquote class="gmail_quote"
style="margin:0pt 0pt 0pt
0.8ex;border-left:1px solid rgb(204, 204,
204);padding-left:1ex"> Several good
records have been published in the
litterature using these<br>
techniques for financial trading
strategies.<br>
</blockquote>
Which ones? References?<br>
<br>
<blockquote class="gmail_quote"
style="margin:0pt 0pt 0pt
0.8ex;border-left:1px solid rgb(204, 204,
204);padding-left:1ex"> There are also
commercial packages (expensive !) which
seem to have achieved<br>
good results.<br>
</blockquote>
Which packages? References again?<br>
<br>
Note that neural network strategies are very
likely to create look-ahead<br>
bias as you develop and test them. You try
something, fail, and try<br>
again on the same data. Unless you are very
careful to reserve a 'pure'<br>
set of instruments and dates that you won't
*ever* touch until you think<br>
you have a 'good' machine learning system,
you're at pretty serious risk<br>
of introducing your look-ahead knowledge
into the system. While this is<br>
true to one degree or another in any
quantitative strategy development,<br>
I think it is a particular risk in
self-adaptive machine learning<br>
methodologies.<br>
<br>
<br>
<blockquote class="gmail_quote"
style="margin:0pt 0pt 0pt
0.8ex;border-left:1px solid rgb(204, 204,
204);padding-left:1ex"> So I feel it could
be nice to share within this group about
the following<br>
subjects :<br>
<br>
- experience using the R packages<br>
- data pre-processing before feeding the
NNs (technical indicators,<br>
wavelets, EMDs, ....)<br>
- which type of NNs are suitable<br>
- how to build and train them<br>
- etc ...<br>
<br>
Thanks to all for sharing within the R
community<br>
</blockquote>
Now, your turn. Bring the community up to
date with your research so<br>
far.<br>
<br>
Regards,<br>
<br>
- Brian<br>
<br>
</blockquote>
<br>
</div>
</div>
<div>
<div>
_______________________________________________<br>
<a moz-do-not-send="true"
href="mailto:R-SIG-Finance@r-project.org"
target="_blank">R-SIG-Finance@r-project.org</a>
mailing list<br>
<a moz-do-not-send="true"
href="https://stat.ethz.ch/mailman/listinfo/r-sig-finance"
target="_blank">https://stat.ethz.ch/mailman/listinfo/r-sig-finance</a><br>
-- Subscriber-posting only. If you want to
post, subscribe first.<br>
-- Also note that this is not the r-help list
where general R questions should go.<br>
</div>
</div>
</blockquote>
</div>
<br>
</div>
</div>
<hr noshade="noshade" size="1">
<p color="#000000" align="left">No virus found in this
message.<br>
Checked by AVG - <a moz-do-not-send="true"
href="http://www.avg.com" target="_blank">www.avg.com</a><br>
Version: 10.0.1392 / Virus Database: 1520/3851 - Release
Date: 08/22/11</p>
</blockquote>
</div>
<br>
_______________________________________________<br>
<a moz-do-not-send="true"
href="mailto:R-SIG-Finance@r-project.org">R-SIG-Finance@r-project.org</a>
mailing list<br>
<a moz-do-not-send="true"
href="https://stat.ethz.ch/mailman/listinfo/r-sig-finance"
target="_blank">https://stat.ethz.ch/mailman/listinfo/r-sig-finance</a><br>
-- Subscriber-posting only. If you want to post, subscribe
first.<br>
-- Also note that this is not the r-help list where general R
questions should go.<br>
</blockquote>
</div>
<br>
<hr noshade="noshade" size="1">
<p class="avgcert" color="#000000" align="left">No virus found in
this message.<br>
Checked by AVG - <a moz-do-not-send="true"
href="http://www.avg.com">www.avg.com</a><br>
Version: 10.0.1392 / Virus Database: 1520/3851 - Release Date:
08/22/11</p>
</blockquote>
</body>
</html>