Newsgroups: comp.ai.neural-nets
From: jimmy@ecowar.demon.co.uk (Jimmy Shadbolt)
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!panix!zip.eecs.umich.edu!newsxfer.itd.umich.edu!gatech!howland.reston.ans.net!pipex!demon!ecowar.demon.co.uk!jimmy
Subject: Re: bond rating: scaling problem 
Distribution: world
References: <19941012093741.BKAMP@pi0192.kub.nl>
Organization: Econostat
Reply-To: jimmy@ecowar.demon.co.uk
X-Newsreader: Simple NEWS 1.90 (ka9q DIS 1.21)
Lines: 26
Date: Fri, 21 Oct 1994 14:43:29 +0000
Message-ID: <782750609snz@ecowar.demon.co.uk>
Sender: usenet@demon.co.uk

In article <19941012093741.BKAMP@pi0192.kub.nl> BKAMP@kub.nl writes:

>Hi,
> 
>I am preparing a bond rating NN (classification). I translated the bond ratings
> 
>to rankings of 0-6. I face the problem that the NN has problems with 
>classifying the higher and lower bond ratings (i.e. 0 and 6).
> 
>I have some experience with NN in stock prediction. In that case you can
>avoid the problem of data (prices) exceeding your training range by 
>translating data to change indicators instead of the actual value.
> 
>Does anyone have suggestions how to use such tactics in a ranked
>classification problem?
> 
>Thanx,
> 
>Bart
>
>

Did you checked/estimated probability density function of your data?

Drago

