Date: Fri, 9 Jan 1998 16:22:00 +0800 Message-Id: <ïßB00> To: bogus From: wmpan@gsc.ict.ac.cn (Wei-Min Pan) Subject: New algorithms for time series predition In-Reply-To: <Main players proving forecasting software>
From: "Wei-Min Pan" <wmpan@gsc.ict.ac.cn>
To: <iif-list@mmu.ac.uk>
Subject: New algorithms for time series predition
Date: Fri, 9 Jan 1998 16:22:00 +0800
This is a multi-part message in MIME format.
------=_NextPart_000_0039_01BD1D1A.B63F7850
Content-Type: text/plain;
charset="gb2312"
Content-Transfer-Encoding: quoted-printable
Dear All,
I am new user of this list. I have just subscribe yesterday. I am very =
glad that i can disccuss problems with the esteemed users of this list.
In the last year, I researched on the bp networks used to predict stock =
price. I found that it can be predict relatively accurately for small =
number prediction. For example, for 50 predictions. And for long use or =
real time use, it is not emperical. I have rearched this problem for =
some time. And now, I have designed a learning algorithm for ANN-based =
predictors. I defined that the predictor which the prediction error does =
not exceed the supper bond of the variation of time series as a validate =
predictor. I proved that the ANN-based predictors with this learning =
algorithm are validate predictor, but that with conventional bp is not a =
validate predictor.Simultanouesly, i designed an algorithm to select the =
size of the delayed input vectors.
If you are interested in the algorthms i design, please contact with me, =
then i will send it to you.
=20
Sincerely,=20
=20
Wei-Min Pan
=20
------=_NextPart_000_0039_01BD1D1A.B63F7850
Content-Type: text/html;
charset="gb2312"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD W3 HTML//EN">
<HTML>
<HEAD>
<META content=3Dtext/html;charset=3Dgb2312 =
http-equiv=3DContent-Type><!DOCTYPE HTML PUBLIC "-//W3C//DTD W3 =
HTML//EN">
<META content=3D'"MSHTML 4.71.1712.3"' name=3DGENERATOR>
</HEAD>
<BODY bgColor=3D#ffffff>
<DIV><FONT face=3D=CB=CE=CC=E5 size=3D2><BR></DIV></FONT>
<DIV>
<DIV><FONT color=3D#000000 size=3D2>Dear All,</FONT></DIV>
<DIV><FONT color=3D#000000 size=3D2>I am new user of this list. I have =
just=20
subscribe yesterday. I am very glad that i can disccuss problems with =
the=20
esteemed users of this list.</FONT></DIV>
<DIV><FONT size=3D2>In the last year, I researched on the bp networks =
used to=20
predict stock price. I found that it can be predict relatively =
accurately for=20
small number prediction. For example, for 50 predictions. And for long =
use or=20
real time use, it is not emperical. I have rearched this problem for =
some=20
time. And now, I have designed a learning algorithm for ANN-based=20
predictors. I defined that the predictor which the prediction error does =
not=20
exceed the supper bond of the variation of time series as a validate =
predictor.=20
I proved that the ANN-based predictors with this learning algorithm are =
validate=20
predictor, but that with conventional bp is not a validate=20
predictor.</FONT><FONT color=3D#000000 size=3D2>Simultanouesly, i =
designed an=20
algorithm to select the size of the delayed input vectors.</FONT></DIV>
<DIV> </DIV>
<DIV><FONT size=3D2>If you are interested in the algorthms i design, =
please=20
contact with me, then i will send it to you.</FONT></DIV>
<DIV><FONT size=3D2></FONT> </DIV>
<DIV><FONT size=3D2>Sincerely, </FONT></DIV>
<DIV><FONT size=3D2></FONT> </DIV>
<DIV><FONT size=3D2>Wei-Min Pan</FONT></DIV>
<DIV><FONT size=3D2></FONT> </DIV></DIV></BODY></HTML>
------=_NextPart_000_0039_01BD1D1A.B63F7850--
This archive was generated by hypermail 2b29 : Thu Jan 06 2000 - 17:14:37 GMT